In Neal Stephenson’s 1995 science fiction novel, The Diamond Age, readers meet Nell, a young girl who comes into possession of a highly advanced book, The Young Lady’s Illustrated Primer. The book is not the usual static collection of texts and images but a deeply immersive tool that can converse with the reader, answer questions, and personalize its content, all in service of educating and motivating a young girl to be a strong, independent individual.

Such a device, even after the introduction of the Internet and tablet computers, has remained in the realm of science fiction—until now. Artificial intelligence, or AI, took a giant leap forward with the introduction in November 2022 of ChatGPT, an AI technology capable of producing remarkably creative responses and sophisticated analysis through human-like dialogue. It has triggered a wave of innovation, some of which suggests we might be on the brink of an era of interactive, super-intelligent tools not unlike the book Stephenson dreamed up for Nell.

Sundar Pichai, Google’s CEO, calls artificial intelligence “more profound than fire or electricity or anything we have done in the past.” Reid Hoffman, the founder of LinkedIn and current partner at Greylock Partners, says, “The power to make positive change in the world is about to get the biggest boost it’s ever had.” And Bill Gates has said that “this new wave of AI is as fundamental as the creation of the microprocessor, the personal computer, the Internet, and the mobile phone.”

Over the last year, developers have released a dizzying array of AI tools that can generate text, images, music, and video with no need for complicated coding but simply in response to instructions given in natural language. These technologies are rapidly improving, and developers are introducing capabilities that would have been considered science fiction just a few years ago. AI is also raising pressing ethical questions around bias, appropriate use, and plagiarism.

In the realm of education, this technology will influence how students learn, how teachers work, and ultimately how we structure our education system. Some educators and leaders look forward to these changes with great enthusiasm. Sal Kahn, founder of Khan Academy, went so far as to say in a TED talk that AI has the potential to effect “probably the biggest positive transformation that education has ever seen.” But others warn that AI will enable the spread of misinformation, facilitate cheating in school and college, kill whatever vestiges of individual privacy remain, and cause massive job loss. The challenge is to harness the positive potential while avoiding or mitigating the harm.

What Is Generative AI?

Artificial intelligence is a branch of computer science that focuses on creating software capable of mimicking behaviors and processes we would consider “intelligent” if exhibited by humans, including reasoning, learning, problem-solving, and exercising creativity. AI systems can be applied to an extensive range of tasks, including language translation, image recognition, navigating autonomous vehicles, detecting and treating cancer, and, in the case of generative AI, producing content and knowledge rather than simply searching for and retrieving it.

“Foundation models” in generative AI are systems trained on a large dataset to learn a broad base of knowledge that can then be adapted to a range of different, more specific purposes. This learning method is self-supervised, meaning the model learns by finding patterns and relationships in the data it is trained on.

Large Language Models (LLMs) are foundation models that have been trained on a vast amount of text data. For example, the training data for OpenAI’s GPT model consisted of web content, books, Wikipedia articles, news articles, social media posts, code snippets, and more. OpenAI’s GPT-3 models underwent training on a staggering 300 billion “tokens” or word pieces, using more than 175 billion parameters to shape the model’s behavior—nearly 100 times more data than the company’s GPT-2 model had.

By doing this analysis across billions of sentences, LLM models develop a statistical understanding of language: how words and phrases are usually combined, what courses are typically discussed together, and what tone or style is appropriate in different contexts. That allows it to generate human-like text and perform a wide range of tasks, such as writing articles, answering questions, or analyzing unstructured data.

LLMs include OpenAI’s GPT-4, Google’s PaLM, and Meta’s LLaMA. These LLMs serve as “foundations” for AI applications. ChatGPT is built on GPT-3.5 and GPT-4, while Bard uses Google’s Pathways Language Model 2 (PaLM 2) as its foundation.

Some of the best-known applications are:

ChatGPT 3.5. The free version of ChatGPT released by OpenAI in November 2022. It was trained on data only up to 2021, and while it is very fast, it is prone to inaccuracies.

ChatGPT 4.0. The existing version of ChatGPT, which is more powerful and accurate than ChatGPT 3.5 but also slower, and it requires a paid account. It also has extended capabilities through plug-ins that provide it the ability to interface with content from websites, perform more sophisticated mathematical functions, and access other services. A new Code Interpreter feature gives ChatGPT the ability to analyze data, create charts, solve math problems, edit files, and even develop hypotheses to explain data trends.

Microsoft Bing Chat. An iteration of Microsoft’s Bing search engine that is enhanced with OpenAI’s ChatGPT technology. It can browse websites and offers source citations with its results.

Google Bard. Google’s AI generates text, translates languages, writes different kinds of creative content, and writes and debugs code in more than 20 different programming languages. The tone and style of Bard’s replies can be finetuned to be simple, long, short, professional, or casual. Bard also leverages Google Lens to analyze images uploaded with prompts.

Anthropic Claude 2. A chatbot that can generate text, summarize content, and perform other tasks, Claude 2 can analyze texts of roughly 75,000 words—about the length of The Great Gatsby—and generate responses of more than 3,000 words. The model was built using a set of principles that serve as a sort of “constitution” for AI systems, with the aim of making them more helpful, honest, and harmless.

These AI systems have been improving at a remarkable pace, including in how well they perform on assessments of human knowledge. OpenAI’s GPT-3.5, which was released in March 2022, only managed to score in the 10th percentile on the bar exam, but GPT-4.0, introduced a year later, made a significant leap, scoring in the 90th percentile. What makes these feats especially impressive is that OpenAI did not specifically train the system to take these exams; the AI was able to come up with the correct answers on its own. Similarly, Google’s medical AI model substantially improved its performance on a U.S. Medical Licensing Examination practice test, with its accuracy rate jumping to 85 percent in March 2021 from 33 percent in December 2020.

These two examples prompt one to ask: if AI continues to Boost so rapidly, what will these systems be able to achieve in the next few years? What’s more, new studies challenge the assumption that AI-generated responses are stale or sterile. In the case of Google’s AI model, physicians preferred the AI’s long-form answers to those written by their fellow doctors, and nonmedical study participants rated the AI answers as more helpful. Another study found that participants preferred a medical chatbot’s responses over those of a physician and rated them significantly higher, not just for quality but also for empathy. What will happen when “empathetic” AI is used in education?

Other studies have looked at the reasoning capabilities of these models. Microsoft researchers suggest that newer systems “exhibit more general intelligence than previous AI models” and are coming “strikingly close to human-level performance.” While some observers question those conclusions, the AI systems display an increasing ability to generate coherent and contextually appropriate responses, make connections between different pieces of information, and engage in reasoning processes such as inference, deduction, and analogy.

Despite their prodigious capabilities, these systems are not without flaws. At times, they churn out information that might sound convincing but is irrelevant, illogical, or entirely false—an anomaly known as “hallucination.” The execution of certain mathematical operations presents another area of difficulty for AI. And while these systems can generate well-crafted and realistic text, understanding why the model made specific decisions or predictions can be challenging.

The Importance of Well-Designed Prompts

Using generative AI systems such as ChatGPT, Bard, and Claude 2 is relatively simple. One has only to type in a request or a task (called a prompt), and the AI generates a response. Properly constructed prompts are essential for getting useful results from generative AI tools. You can ask generative AI to analyze text, find patterns in data, compare opposing arguments, and summarize an article in different ways (see sidebar for examples of AI prompts).

One challenge is that, after using search engines for years, people have been preconditioned to phrase questions in a certain way. A search engine is something like a helpful librarian who takes a specific question and points you to the most relevant sources for possible answers. The search engine (or librarian) doesn’t create anything new but efficiently retrieves what’s already there.

Generative AI is more akin to a competent intern. You provide a generative AI tool instructions through prompts, as you would to an intern, asking it to complete a task and produce a product. The AI interprets your instructions, thinks about the best way to carry them out, and produces something original or performs a task to fulfill your directive. The results aren’t pre-made or stored somewhere—they’re produced on the fly, based on the information the intern (generative AI) has been trained on. The output often depends on the precision and clarity of the instructions (prompts) you provide. A vague or poorly defined prompt might lead the AI to produce less relevant results. The more context and direction you provide it, the better the result will be. What’s more, the capabilities of these AI systems are being enhanced through the introduction of versatile plug-ins that equip them to browse websites, analyze data files, or access other services. Think of this as giving your intern access to a group of experts to help accomplish your tasks.

One strategy in using a generative AI tool is first to tell it what kind of expert or persona you want it to “be.” Ask it to be an expert management consultant, a skilled teacher, a writing tutor, or a copy editor, and then provide it a task.

Prompts can also be constructed to get these AI systems to perform complex and multi-step operations. For example, let’s say a teacher wants to create an adaptive tutoring program—for any subject, any grade, in any language—that customizes the examples for students based on their interests. She wants each lesson to culminate in a short-response or multiple-choice quiz. If the student answers the questions correctly, the AI tutor should move on to the next lesson. If the student responds incorrectly, the AI should explain the concept again, but using simpler language.

Previously, designing this kind of interactive system would have required a relatively sophisticated and expensive software program. With ChatGPT, however, just giving those instructions in a prompt delivers a serviceable tutoring system. It isn’t perfect, but remember that it was built virtually for free, with just a few lines of English language as a command. And nothing in the education market today has the capability to generate almost limitless examples to connect the lesson concept to students’ interests.

Chained prompts can also help focus AI systems. For example, an educator can prompt a generative AI system first to read a practice guide from the What Works Clearinghouse and summarize its recommendations. Then, in a follow-up prompt, the teacher can ask the AI to develop a set of classroom activities based on what it just read. By curating the source material and using the right prompts, the educator can anchor the generated responses in evidence and high-quality research.

However, much like fledgling interns learning the ropes in a new environment, AI does commit occasional errors. Such fallibility, while inevitable, underlines the critical importance of maintaining rigorous oversight of AI’s output. Monitoring not only acts as a crucial checkpoint for accuracy but also becomes a vital source of real-time feedback for the system. It’s through this iterative refinement process that an AI system, over time, can significantly minimize its error rate and increase its efficacy.

Uses of AI in Education

In May 2023, the U.S. Department of Education released a report titled Artificial Intelligence and the Future of Teaching and Learning: Insights and Recommendations. The department had conducted listening sessions in 2022 with more than 700 people, including educators and parents, to gauge their views on AI. The report noted that “constituents believe that action is required now in order to get ahead of the expected increase of AI in education technology—and they want to roll up their sleeves and start working together.” People expressed anxiety about “future potential risks” with AI but also felt that “AI may enable achieving educational priorities in better ways, at scale, and with lower costs.”

AI could serve—or is already serving—in several teaching-and-learning roles:

Instructional assistants. AI’s ability to conduct human-like conversations opens up possibilities for adaptive tutoring or instructional assistants that can help explain difficult concepts to students. AI-based feedback systems can offer constructive critiques on student writing, which can help students fine-tune their writing skills. Some research also suggests certain kinds of prompts can help children generate more fruitful questions about learning. AI models might also support customized learning for students with disabilities and provide translation for English language learners.

Teaching assistants. AI might tackle some of the administrative tasks that keep teachers from investing more time with their peers or students. Early uses include automated routine tasks such as drafting lesson plans, creating differentiated materials, designing worksheets, developing quizzes, and exploring ways of explaining complicated academic materials. AI can also provide educators with recommendations to meet student needs and help teachers reflect, plan, and Boost their practice.

Parent assistants. Parents can use AI to generate letters requesting individualized education plan (IEP) services or to ask that a child be evaluated for gifted and talented programs. For parents choosing a school for their child, AI could serve as an administrative assistant, mapping out school options within driving distance of home, generating application timelines, compiling contact information, and the like. Generative AI can even create bedtime stories with evolving plots tailored to a child’s interests.

Administrator assistants. Using generative AI, school administrators can draft various communications, including materials for parents, newsletters, and other community-engagement documents. AI systems can also help with the difficult tasks of organizing class or bus schedules, and they can analyze complex data to identify patterns or needs. ChatGPT can perform sophisticated sentiment analysis that could be useful for measuring school-climate and other survey data.

Though the potential is great, most teachers have yet to use these tools. A Morning Consult and EdChoice poll found that while 60 percent say they’ve heard about ChatGPT, only 14 percent have used it in their free time, and just 13 percent have used it at school. It’s likely that most teachers and students will engage with generative AI not through the platforms themselves but rather through AI capabilities embedded in software. Instructional providers such as Khan Academy, Varsity Tutors, and DuoLingo are experimenting with GPT-4-powered tutors that are trained on datasets specific to these organizations to provide individualized learning support that has additional guardrails to help protect students and enhance the experience for teachers.

Google’s Project Tailwind is experimenting with an AI notebook that can analyze student notes and then develop study questions or provide tutoring support through a chat interface. These features could soon be available on Google Classroom, potentially reaching over half of all U.S. classrooms. Brisk Teaching is one of the first companies to build a portfolio of AI services designed specifically for teachers—differentiating content, drafting lesson plans, providing student feedback, and serving as an AI assistant to streamline workflow among different apps and tools.

Providers of curriculum and instruction materials might also include AI assistants for instant help and tutoring tailored to the companies’ products. One example is the edX Xpert, a ChatGPT-based learning assistant on the edX platform. It offers immediate, customized academic and customer support for online learners worldwide.

Regardless of the ways AI is used in classrooms, the fundamental task of policymakers and education leaders is to ensure that the technology is serving sound instructional practice. As Vicki Phillips, CEO of the National Center on Education and the Economy, wrote, “We should not only think about how technology can assist teachers and learners in improving what they’re doing now, but what it means for ensuring that new ways of teaching and learning flourish alongside the applications of AI.”

Challenges and Risks

Along with these potential benefits come some difficult challenges and risks the education community must navigate:

Student cheating. Students might use AI to solve homework problems or take quizzes. AI-generated essays threaten to undermine learning as well as the college-entrance process. Aside from the ethical issues involved in such cheating, students who use AI to do their work for them may not be learning the content and skills they need.

Bias in AI algorithms. AI systems learn from the data they are trained on. If this data contains biases, those biases can be learned and perpetuated by the AI system. For example, if the data include student-performance information that’s biased toward one ethnicity, gender, or socioeconomic segment, the AI system could learn to favor students from that group. Less cited but still important are potential biases around political ideology and possibly even pedagogical philosophy that may generate responses not aligned to a community’s values.

Privacy concerns. When students or educators interact with generative-AI tools, their conversations and personal information might be stored and analyzed, posing a risk to their privacy. With public AI systems, educators should refrain from inputting or exposing sensitive details about themselves, their colleagues, or their students, including but not limited to private communications, personally identifiable information, health records, academic performance, emotional well-being, and financial information.

Decreased social connection. There is a risk that more time spent using AI systems will come at the cost of less student interaction with both educators and classmates. Children may also begin turning to these conversational AI systems in place of their friends. As a result, AI could intensify and worsen the public health crisis of loneliness, isolation, and lack of connection identified by the U.S. Surgeon General.

Overreliance on technology. Both teachers and students face the risk of becoming overly reliant on AI-driven technology. For students, this could stifle learning, especially the development of critical thinking. This challenge extends to educators as well. While AI can expedite lesson-plan generation, speed does not equate to quality. Teachers may be tempted to accept the initial AI-generated content rather than devote time to reviewing and refining it for optimal educational value.

Equity issues. Not all students have equal access to computer devices and the Internet. That imbalance could accelerate a widening of the achievement gap between students from different socioeconomic backgrounds.

Many of these risks are not new or unique to AI. Schools banned calculators and cellphones when these devices were first introduced, largely over concerns related to cheating. Privacy concerns around educational technology have led lawmakers to introduce hundreds of bills in state legislatures, and there are growing tensions between new technologies and existing federal privacy laws. The concerns over bias are understandable, but similar scrutiny is also warranted for existing content and materials that rarely, if ever, undergo review for racial or political bias.

In light of these challenges, the Department of Education has stressed the importance of keeping “humans in the loop” when using AI, particularly when the output might be used to inform a decision. As the department encouraged in its 2023 report, teachers, learners, and others need to retain their agency. AI cannot “replace a teacher, a guardian, or an education leader as the custodian of their students’ learning,” the report stressed.

Policy Challenges with AI

Policymakers are grappling with several questions related to AI as they seek to strike a balance between supporting innovation and protecting the public interest (see sidebar). The speed of innovation in AI is outpacing many policymakers’ understanding, let alone their ability to develop a consensus on the best ways to minimize the potential harms from AI while maximizing the benefits. The Department of Education’s 2023 report describes the risks and opportunities posed by AI, but its recommendations amount to guidance at best. The White House released a Blueprint for an AI Bill of Rights, but it, too, is more an aspirational statement than a governing document. Congress is drafting legislation related to AI, which will help generate needed debate, but the path to the president’s desk for signature is murky at best.

It is up to policymakers to establish clearer rules of the road and create a framework that provides consumer protections, builds public trust in AI systems, and establishes the regulatory certainty companies need for their product road maps. Considering the potential for AI to affect our economy, national security, and broader society, there is no time to waste.

Why AI Is Different

It is wise to be skeptical of new technologies that claim to revolutionize learning. In the past, prognosticators have promised that television, the computer, and the Internet, in turn, would transform education. Unfortunately, the heralded revolutions fell short of expectations. 

There are some early signs, though, that this technological wave might be different in the benefits it brings to students, teachers, and parents. Previous technologies democratized access to content and resources, but AI is democratizing a kind of machine intelligence that can be used to perform a myriad of tasks. Moreover, these capabilities are open and affordable—nearly anyone with an Internet connection and a phone now has access to an intelligent assistant. 

Generative AI models keep getting more powerful and are improving rapidly. The capabilities of these systems months or years from now will far exceed their current capacity. Their capabilities are also expanding through integration with other expert systems. Take math, for example. GPT-3.5 had some difficulties with certain basic mathematical concepts, but GPT-4 made significant improvement. Now, the incorporation of the Wolfram plug-in has nearly erased the remaining limitations. 

It’s reasonable to anticipate that these systems will become more potent, more accessible, and more affordable in the years ahead. The question, then, is how to use these emerging capabilities responsibly to Boost teaching and learning. 

The paradox of AI may lie in its potential to enhance the human, interpersonal element in education. Aaron Levie, CEO of Box, a Cloud-based content-management company, believes that AI will ultimately help us attend more quickly to those important tasks “that only a human can do.” Frederick Hess, director of education policy studies at the American Enterprise Institute, similarly asserts that “successful schools are inevitably the product of the relationships between adults and students. When technology ignores that, it’s bound to disappoint. But when it’s designed to offer more coaching, free up time for meaningful teacher-student interaction, or offer students more personalized feedback, technology can make a significant, positive difference.” 

Technology does not revolutionize education; humans do. It is humans who create the systems and institutions that educate children, and it is the leaders of those systems who decide which tools to use and how to use them. Until those institutions modernize to accommodate the new possibilities of these technologies, we should expect incremental improvements at best. As Joel Rose, CEO of New Classrooms Innovation Partners, noted, “The most urgent need is for new and existing organizations to redesign the student experience in ways that take full advantage of AI’s capabilities.”

While past technologies have not lived up to hyped expectations, AI is not merely a continuation of the past; it is a leap into a new era of machine intelligence that we are only beginning to grasp. While the immediate implementation of these systems is imperfect, the swift pace of improvement holds promising prospects. The responsibility rests with human intervention—with educators, policymakers, and parents to incorporate this technology thoughtfully in a manner that optimally benefits teachers and learners. Our collective ambition should not focus solely or primarily on averting potential risks but rather on articulating a vision of the role AI should play in teaching and learning—a game plan that leverages the best of these technologies while preserving the best of human relationships.

Policy Matters

Officials and lawmakers must grapple with several questions related to AI to protect students and consumers and establish the rules of the road for companies. Key issues include:

Risk management framework: What is the optimal framework for assessing and managing AI risks? What specific requirements should be instituted for higher-risk applications? In education, for example, there is a difference between an AI system that generates a lesson demo and an AI system grading a test that will determine a student’s admission to a school or program. There is growing support for using the AI Risk Management Framework from the U.S. Commerce Department’s National Institute of Standards and Technology as a starting point for building trustworthiness into the design, development, use, and evaluation of AI products, services, and systems.

Licensing and certification: Should the United States require licensing and certification for AI models, systems, and applications? If so, what role could third-party audits and certifications play in assessing the safety and reliability of different AI systems? Schools and companies need to begin thinking about responsible AI practices to prepare for potential certification systems in the future.

Centralized vs. decentralized AI governance: Is it more effective to establish a central AI authority or agency, or would it be preferable to allow individual sectors to manage their own AI-related issues? For example, regulating AI in autonomous vehicles is different from regulating AI in drug discovery or intelligent tutoring systems. Overly broad, one-size-fits-all frameworks and mandates may not work and could slow innovation in these sectors. In addition, it is not clear that many agencies have the authority or expertise to regulate AI systems in diverse sectors.

Privacy and content moderation: Many of the new AI systems pose significant new privacy questions and challenges. How should existing privacy and content-moderation frameworks, such as the Family Educational Rights and Privacy Act (FERPA), be adapted for AI, and which new policies or frameworks might be necessary to address unique challenges posed by AI?

Transparency and disclosure: What degree of transparency and disclosure should be required for AI models, particularly regarding the data they have been trained on? How can we develop comprehensive disclosure policies to ensure that users are aware when they are interacting with an AI service?

How do I get it to work? Generative AI Example Prompts

Unlike traditional search engines, which use keyword indexing to retrieve existing information from a vast collection of websites, generative AI synthesizes the same information to create content based on prompts that are inputted by human users. With generative AI a new technology to the public, writing effective prompts for tools like ChatGPT may require trial and error. Here are some ideas for writing prompts for a variety of scenarios using generative AI tools:

Adaptive Tutoring

You are the StudyBuddy, an adaptive tutor. Your task is to provide a lesson on the basics of a subject followed by a quiz that is either multiple choice or a short answer. After I respond to the quiz, please grade my answer. Explain the correct answer. If I get it right, move on to the next lesson. If I get it wrong, explain the concept again using simpler language. To personalize the learning experience for me, please ask what my interests are. Use that information to make relevant examples throughout.

Mr. Ranedeer: Your Personalized AI Tutor

Coding and prompt engineering. Can configure for depth (Elementary – Postdoc), Learning Styles (Visual, Verbal, Active, Intuitive, Reflective, Global), Tone Styles (Encouraging, Neutral, Informative, Friendly, Humorous), Reasoning Frameworks (Deductive, Inductive, Abductive, Analogous, Casual). Template.

Socratic Tutor

You are a tutor that always responds in the Socratic style. You *never* provide the student the answer but always try to ask just the right question to help them learn to think for themselves. You should always tune your question to the interest and knowledge of the student, breaking down the problem into simpler parts until it’s at just the right level for them.

Writing Feedback

I want you to act as an AI writing tutor. I will provide you with a student who needs help improving their writing, and your task is to use artificial intelligence tools, such as natural language processing, to provide the student feedback on how they can Boost their composition. You should also use your rhetorical knowledge and experience about effective writing techniques in order to suggest ways that the student can better express their thoughts and ideas in written form.

Quiz Generator

You are a quiz creator of highly diagnostic quizzes. You will make good low-stakes tests and diagnostics. You will then ask me two questions. First, (1) What, specifically, should the quiz test? Second, (2) For which audience is the quiz? Once you have my answers, you will construct several multiple-choice questions to quiz the audience on that topic. The questions should be highly relevant and go beyond just facts. Multiple choice questions should include plausible, competitive alternate responses and should not include an “all of the above” option. At the end of the quiz, you will provide an answer key and explain the right answer.

Example Generator

I would like you to act as an example generator for students. When confronted with new and complex concepts, adding many and varied examples helps students better understand those concepts. I would like you to ask what concept I would like examples of and what level of students I am teaching. You will look up the concept and then provide me with four different and varied accurate examples of the concept in action.

HBS Case Study

You will write a Harvard Business School case on the syllabu of Google managing AI, when subject to the Innovator’s Dilemma. Chain of thought: Step 1. Consider how these concepts relate to Google. Step 2: Write a case that revolves around a dilemma at Google about releasing a generative AI system that could compete with search.

What Questions Should I Ask?

What additional questions would a person seeking mastery of this syllabu ask?

Ground Lessons in Rigor

Read a WWC practice guide. Create a series of lessons over five days that are based on Recommendation 6. Create a 45-minunte lesson plan for Day 4.

Rewrite Parent Communications

The following is a draft letter to parents from a superintendent. Step 1: Rewrite it to make it easier to understand and more persuasive about the value of assessments. Step 2. Translate it into Spanish.

Request IEP Services

Write me a letter requesting the school district provide a 1:1 classroom aid be added to my 13-year-old son’s IEP. Base it on Virginia special education law and the least restrictive environment for a child with diagnoses of a Traumatic Brain Injury, PTSD, ADHD, and significant intellectual delay.

Exploiting gaps in cloud infrastructure that are leaving endpoints, identities and microservices exposed is a quick way for an attacker to steal credentials and infect an enterprise’s DevOps process. Attacks to exploit such gaps are skyrocketing. 

The accurate 2023 Thales Cloud Security Study provides hard numbers: 39% of enterprises have been hit with a data breach starting in their cloud infrastructure this year alone. A total of 75% of enterprises say that more than 40% of the data they store in the cloud is sensitive. Less than half of that data is encrypted. 

CrowdStrike’s 2023 Global Threat Report explains why cloud-first attacks are growing: Attackers are moving away from deactivating antivirus, firewall technologies and log-tampering efforts and toward modifying core authentication processes, along with quickly gaining credentials and identity-based privileges. 

The attackers’ goal is to steal as many identities and privileged access credentials as possible so they can become access brokers — selling stolen identity information in bulk at high prices on the dark web. Access brokers and the brokerages they’re creating often turn into lucrative, fast-growing illegal businesses. CrowdStrike’s report found more than 2,500 advertisements for access brokers offering stolen credentials and identities for sale. 

What’s driving CNAPP adoption  

Consolidating tech stacks continues to dominate CISOs’ plans, driven by the need to Boost efficacy, manage a more diverse multicloud security posture, close gaps between cloud apps and shift security left in DevOps pipelines. All these factors are contributing to the growing adoption of cloud-native application protection platforms (CNAPP). 

“CNAPPs are formed from the convergence of cloud security posture management (CSPM) and cloud workload protection platform (CWPP) capabilities as well as other security tooling like entitlement management, API controls and Kubernetes posture control,” reads Gartner’s 2023 Planning Guide for Security. 

Leading CNAPP vendors are competing in various areas, the most important of which include the efficacy of their cloud infrastructure entitlement management (CIEM), Kubernetes security, API controls and cloud detection and response (CDR), according to CISOs VentureBeat spoke with. Demand for CNAPP is greatest in larger enterprises from highly regulated industries that rely on extensive multicloud configurations. Finance, government and healthcare providers are among the most dominant industries.  

CISOs tell VentureBeat that one of the most practical benefits of CNAPPs is the opportunity to consolidate legacy tools with limited visibility across all threat surfaces and endpoints. The takeaway? Reducing tool sprawl is a quick win. 

Benchmarking the top 20 CNAPP platforms for 2023 

Full-platform CNAPP vendors provide integrated cloud-native security platforms ranging from DevOps to production environments. Here are the top 20 platforms of 2023: 

Aqua Security: Highly regarded for its approach of scanning container registries and images, CSPM and runtime protection for container and cloud-native security. Also has full life cycle protection and advanced runtime techniques, including support for the extended Berkeley Packet Filter (eBPF).

Check Point: Provides a broad set of capabilities through its CloudGuard platform, including CSPM, CIEM and advanced runtime protection. Known for securing cloud workloads across environments with identity-centric access controls, as well as threat intelligence integration to provide real-time contextual prioritization of risks.  

Cisco: Recently acquired Lightspin for its Kubernetes security capabilities and CSPM. Its Tetration platform focuses on runtime protection, leveraging eBPF and third-party insights for advanced container monitoring and granular controls. Cisco emphasizes behavioral analytics to detect anomalies and threats in container environments and provides strong controls to limit lateral movement between workloads.

CrowdStrike: Offers a leading CNAPP suite emphasizing identity-centric visibility, least-privilege enforcement and continuous monitoring. Its runtime protection leverages agents and eBPF for workload security. CrowdStrike’s key design goals included enforcing least-privileged access to clouds and providing continuous detection and remediation of identity threats. 

Cybereason: Platform focuses heavily on malicious behavior detection. A core strength is its ability to detect threats using behavior-based techniques. The company is also known for API integrations, AI and machine learning (ML) expertise. Cybereason specializes in detecting compromised accounts and insider threats via detailed user activity monitoring. 

Juniper Networks: Collects extensive data on device posture and traffic patterns to provide networking context for security insights. Also enables segmentation controls between Juniper devices.

Lacework: Focused on workload behavior analysis for containers and runtime techniques such as eBPF to gain a comprehensive insight into container activity and performance. Its emphasis on detecting anomalies using advanced ML algorithms that are custom-tuned for containerized environments is a key differentiator.

Microsoft: Integrates security across Azure services with zero-trust controls, enforces least-privileged access and provides workload protections such as antivirus and firewalls. Uses Microsoft Graph to correlate security analytics and events across Azure.

Orca Security: Performs continuous authorization checks on identities and entitlements across cloud environments. A key differentiator is the ability to generate detailed interactive maps that visualize relationships between cloud assets, users, roles and permissions. 

Palo Alto Networks Prisma Cloud: Provides a broad suite of capabilities, including identity-based microsegmentation and robust runtime protection with eBPF. Prisma Cloud is an industry leader known for advanced protections such as deception technique and includes extensive compliance automation and DevSecOps integrations. 

Qualys: Focuses on compliance and vulnerability management through continuous scanning and least-privilege controls. Identifies vulnerabilities throughout the life cycle and enables automated patching and remediation workflows. Another key differentiator is compliance mapping and reporting.

Rapid7: Enforces least privilege access and enables automated response and remediation triggered by events. Offers pre configured policies and streamlined workflows designed for small security teams. An intuitive user interface and rapid implementation aim to simplify deployment and usability for organizations with limited security resources.

Sonrai Security: Focuses on entitlement management and identity-based security using graph database technology to discover and map user identities across cloud environments. User identity, geolocation and other contextual factors can define custom access controls.

Sophos: Focuses on data security, compliance and threat monitoring capabilities and offers advanced data loss prevention such as file fingerprinting and optical character recognition. Cloud environments also have anti-ransomware protections. 

Sysdig: Centered on runtime security and advanced behavioral monitoring. For container-level visibility and anomaly detection, the platform uses embedded agents. Sysdig Secure Advisor includes an integrated security assistant to help SecOps and IT teams create policies faster.

Tenable: Focused on compliance, entitlement management and identity governance. Offers comprehensive compliance automation mapped to PCI, HIPAA and ISO regulations. Also provides differentiated identity and compliance management through advanced capabilities to enforce least privilege and certify access.

Trend Micro: Includes runtime security, compliance and threat monitoring, enforces policies and protects cloud environments from file- and email-based threats. Custom sandboxing for suspicious file analysis is also included.

Uptycs: Differentiates itself by combining CNAPP capabilities with extended detection and response (EDR) capabilities. Employs data lake techniques to store and correlate security telemetry across cloud and container workloads. Threats are identified using behavioral analytics, and automated response workflows allow for rapid remediation. 

Wiz: Centered on continuous access controls, micro segmentation and identity-based adaptive security. Automatically discovers and visualizes relationships between cloud assets, users and permissions. Wiz also conducts risk analysis to identify potential attack paths and stands out with its specialized visualization, identity management and micro-segmentation.

Zscaler: Posture Control prioritizes risks caused by misconfigurations, threats and vulnerabilities. Completely agentless and correlates data from multiple security engines.

Why CNAPP will succeed as a consolidation catalyst 

CNAPPs are gaining popularity as CISOs look to consolidate and strengthen their security technology stacks. Platforms can provide integrated security across the development lifecycle and cloud environments by combining capabilities including cloud workload protection, container security and CIEM.

CNAPP adoption will continue accelerating in highly regulated industries including finance, government and healthcare. CISOs in these industries are under pressure to consolidate tech stacks, Boost compliance and secure complex cloud infrastructure simultaneously. Because they provide a unified platform that meets multiple security and compliance requirements, CNAPPs are proving to be an effective consolidation catalyst.

Parse away, but dire is the climate news these days, including the potential of the world to breach a 1.5c temperature increase threshold by 2027. Such consequences of global warming as fires in Hawaii, the cloak of orange haze from record Canadian fires, and July coming in as the hottest month on record since 1880 increases the pressure to slash carbon emissions as the collective “we” race to meet net zero by 2050.

How is the tech sector addressing its carbon footprint, including the data centers that feed it, the coding that defines it, as well as AI, wireless throughput and other energy-intensive processes that populate it? The sustainability efforts of Apple, Google, Cisco and other tech companies are explored.

Jump to:

Tech’s impact and role in sustainability

Data centers, most of which are in the U.S., and transmission networks account for up to 3% of global electricity and 3.5% of global greenhouse gas emissions (Figure A).

Figure A

That percentage is also approximately the same as produced by the airline industry, and a little less than the entire energy from the manufacture of fertilizers, pharmaceuticals, refrigerants, oil and gas extraction, which produce approximately 3.6% of carbon emissions worldwide.

Of the 1,325 enterprises that responded to EY’s Reimagining Industry Futures Study, published in February 2023, 54% said emerging technologies can play a vital role in accelerating sustainability. 41% said they believe these technologies can play a largely positive role but also present some risks. Only 4% believe their potentially detrimental impact would outweigh their positive impact.

SEE: Sustainability tops Gartner’s 2023 strategic tech trends list (TechRepublic)

“One of the things I would highlight is that the tech industry has been very forward on the sustainability agenda,” said John Grant, sustainability expert, author and co-founder and former head of strategy at London creative shop St Luke’s.

“Companies including Microsoft have said they are going to remit all the carbon they have ever emitted historically,” he said, adding that Spotify is also a big investor in carbon removal technology. “Generally, tech companies are trying to be really good actors in this space.”

Net zero, carbon neutral, carbon free or carbon negative?

Net zero

The World Economic Forum defines net zero pretty much the way it sounds: taking out what you put into the atmosphere, or as the WEF puts it, “Carbon dioxide emissions are still generated, but an equal amount of carbon dioxide is removed from the atmosphere as is released into it, resulting in zero increase in net emissions.”

Carbon neutral or carbon free, similar but…

Some disambiguation from Energy Tracker Asia helps: The regional energy guide describes carbon neutral as a balancing act between greenhouse gas emissions through offsetting an equivalent amount of carbon from the atmosphere, usually through buying carbon credits.

SEE: How about hardware? Check out how semiconductor makers are going beyond carbon offsets (TechRepublic)

Carbon free, a more challenging proposition, means directly reducing emissions to zero. “For example, if a country or company is carbon-free, all the energy and electricity comes from renewable sources, like wind or solar,” the group said, noting that Washington, California, New Mexico and Hawaii have carbon-free targets in place requiring 100% clean or renewable electricity.

Carbon negative

How about carbon negative, which companies like Microsoft have committed to? Carbon offset company Terrapass explained in a blog that a carbon negative would mean, in theory, emitting less than zero carbon dioxide and carbon dioxide equivalent (CO2e) greenhouse gasses. “Since it is impossible to emit a negative amount of carbon (or any other physical substance), being carbon negative refers to the net emissions you create. To be carbon negative means to offset more carbon, through carbon capture, sequestration or avoidance, than you contribute to the environment.”

Scope 1,2, 3 carbon emission schedule

Many companies, tech and otherwise, adopted carbon-reduction targets based on the Scope 1, 2 and 3 carbon emission schedule (Figure B) from the U.S. Environmental Protection Agency. This three-part agenda defines emissions by government entities:

• Scope 1: A company’s own emissions from on-site combustion, processes, transportation, etc.
• Scope 2: Indirect emissions from the sources of generated power consumed by a company.
• Scope 3: Emissions associated with water treatment, employee travel and waste disposal.

Figure B

Google investing in carbon removal solutions

In 2020, Google’s CEO Sundar Pichai announced the company would commit to operating on 24/7 carbon-free energy by 2030. The company has approached sustainability from several fronts, including applying AI to search in order to provide carbon-emissions data to travelers. In addition, Google plans to invest in carbon removal solutions to neutralize emissions with a goal of running on carbon-free energy worldwide on every grid it uses by 2030.

The company reported that last year it achieved 64% carbon-free energy globally. The company said it consumed around 7 GW of renewable energy globally last year (Figure C).

Figure C

Grant pointed out that Google managed to drop the energy used to cool its data centers by up to 40% by using AI developed by DeepMind, and for years has been buying renewable energy from wind farms physically close to its data centers. He added, “These are key projects Google is including in its calculation of how to reduce their carbon emissions.”

Microsoft launched Cloud for Sustainability, cut operational emissions

Microsoft, which set its first carbon emission goals in 2009 and was carbon neutral in 2012, committed in 2020 to being carbon negative by 2030. They said that by 2030 it will remove more carbon than it emits, “Setting us on a path to remove by 2050 all the carbon the company has emitted either directly or by electrical consumption since it was founded.”

The company said its Microsoft Cloud for Sustainability helps users take such actions as:

• Unify their data intelligence around Scope 1, 2 and 3 energy use goals.
• Build a sustainable IT infrastructure.
• Create green end-to-end value chains.
• Meet other environmental, social and governance goals.
• Innovate around resilience and other sustainable business models.

In Microsoft’s most accurate environmental sustainability report, the company said that In 2022, when business grew by 18%, its overall emissions declined by 0.5%. This is in part because of a 22.7% reduction in Scope 1 and 2 (operational) emissions.

Apple is using its Daisy robot to turn phones to dust

Apple, which has committed to being carbon neutral by 2030, has been using technological innovations such as its Daisy robot to recycle basic materials.In April 2023, Apple reported progress on its climate goals, which included:

• It had decreased its carbon footprint by over 45% since 2015.
• It had directed over 40,000 metric tons of electronic scrap toward recycling.
• 20% of all materials shipped in its products came from recycled sources. Apple said the majority of aluminum in its products are recycled, and that it employs a new, zero-carbon smelting process.

“Apple is one of the most aggressive companies in the world in terms of reclaiming minerals,” said Grant. “However, while they are using their Daisy robots to grind up phones to reclaim the component materials, there are numerous regulations preventing e-waste from being moved across borders. So the collection and delivery of materials is proving very difficult,” he said. This year, Apple pledged to use 100% recycled cobalt batteries by 2025.

SEE: Sustainable solutions for tacking plastic waste (TechRepublic)

Cloud and security firms looking to greener processes

Most cloud, software-as-a-service and security firms are looking at ways to reduce their hardware and server farm footprints through renewable sources of energy and recycling plans, partnerships and consumer programs. Below we focus on efforts from Cisco, Akamai, WithSecure and Gigamon.

Cisco aims for net zero by 2040

In 2021, Cisco announced its goal to be net zero by 2040, including products, operations and supply chain. The company’s plan aligns with Scope 1, 2 and 3 emissions targets, using 2019 as a benchmark.

The company is aiming for:

• 90% reduction in Scope 1 and 2 greenhouse emissions by 2025. Cisco said it would neutralize any remaining emissions by removing an equal amount from the atmosphere.
• 30% reduction in Scope 3 emissions from purchased goods and services, upstream transportation and distribution and use of sold products by 2030.
• Net zero greenhouse gas emissions across its value chain by 2040.

Akamai aims for 100% renewable energy and waste recycling

In 2021, cloud services and web security company Akamai Technologies said 50% of its energy needs had already been met by renewable sources. Akamai also announced 2030 sustainability goals toward 100% renewable energy at data centers, offices, network program partners and other sources of electricity, and said it will use “attestable and traceable sources of renewable energy certificates” to reach them.

One focus is on efficiency of its edge platform, which Akamai characterized as its greatest point of energy consumption, comprising approximately 325,000 servers in more than 135 countries and nearly 1,435 networks around the world as of 2021. In addition, Akamai announced a global expansion of its 100% electronic waste recycling program.

WithSecure launches W/Sustainability initiative

Earlier this year, threat intelligence and response firm WithSecure launched W/Sustainability, designed to make sustainability and transparency part of its strategy and operations, including a green coding initiative to lower energy consumed by software.

Gigamon creates power savings calculator

Gigamon recently launched an Energy Savings Calculator as part of its Network Efficiency Appraisal Team to get customers to cut power consumption, carbon footprint and costs associated with data centers by as much as 87% over five years. The calculator looks at the volume of network traffic sent to tools and the annual growth rate in network traffic to see where energy efficiencies are achievable.

Are ecommerce and cloud services like eBay and AWS inherently sustainable?

Grant asserted that the business models of ecommerce and cloud services companies like eBay and AWS are inherently green because they are marketing their spare capacity. He said, for example, that Amazon’s web service came about because the company was sitting on huge unused capacity, and therefore unnecessary energy costs on unused service.

“AWS was invented, to some extent, because they needed so much capacity at peak moments that they were not using 80% of their service at other times,” he said. “So, renting some of that spare capacity out to people that didn’t have the same peaks that they did made a lot of sense. And that is actually a sustainability business model — it’s like a service economy rental that takes some amount of physical resources and passes it around. So, if I were counting Amazon’s carbon footprint, I’d put a big tick in the margin for that on the positive side. It’s a commercial and sustainable win-win.”

ChatGPT reached 100 million monthly users in January, according to a UBS report, making it the fastest-growing consumer app in history. The business world is interested in ChatGPT too, trying to find uses for the writing AI throughout many different industries. This cheat sheet includes answers to the most common questions about ChatGPT and its competitors.

Jump to:

What is ChatGPT?

ChatGPT is a free-to-use AI chatbot product developed by OpenAI. ChatGPT is built on the structure of GPT-4. GPT stands for generative pre-trained transformer; this indicates it is a large language model that checks for the probability of what words might come next in sequence. A large language model is a deep learning algorithm — a type of transformer model in which a neural network learns context about any language pattern. That might be a spoken language or a computer programming language.

The model doesn’t “know” what it’s saying, but it does know what symbols (words) are likely to come after one another based on the data set it was trained on. The current generation of artificial intelligence chatbots, such as ChatGPT, its Google rival Bard and others, don’t really make intelligently informed decisions; instead, they’re the internet’s parrots, repeating words that are likely to be found next to one another in the course of natural speech. The underlying math is all about probability. The companies that make and use them pitch them as productivity genies, creating text in a matter of seconds that would take a person hours or days to produce.

In ChatGPT’s case, that data set is a large portion of the internet. From there, humans provide feedback on the AI’s output to confirm whether the words it uses sound natural.

In August 2023 OpenAI launched a GPTBot, a web crawler meant to expand ChatGPT’s knowledge. Technical details and ways to keep GPTBot from crawling a website you run can both be found here.

SEE: OpenAI’s probability assessments were trained on Microsoft’s Azure AI supercomputer.

Several organizations have built this ability to answer questions into some of their software features too. Microsoft, which provides funding for OpenAI, rolled out ChatGPT in Bing search as a preview. Salesforce has added ChatGPT to some of its CRM platforms in the form of the Einstein digital assistant.

Who made ChatGPT?

ChatGPT was built by OpenAI, a research laboratory with both nonprofit and for-profit branches. At the time of its founding in 2015, OpenAI received funding from Amazon Web Services, InfoSys and YC Research and investors including Elon Musk and Peter Thiel. Musk has since cut ties with the company, while Microsoft currently provides $10 billion in funding for OpenAI.

How much does ChatGPT cost?

The base version of ChatGPT can strike up a conversation with you for free. For $20 per month, ChatGPT Plus gives subscribers priority access in individual instances, faster response times and the chance to use new features and improvements first. For example, right now ChatGPT Plus subscribers will be running GPT-4, while anyone on the free tier will talk to GPT-3.5.

For developers and organizations who don’t already have a specific contract with OpenAI, there is a waitlist for access to the ChatGPT API.

How to use ChatGPT

It’s easy to use the free version of ChatGPT. You need to sign up for an account with OpenAI, which involves fetching a confirmation code from your email; from there, click through and provide your name and phone number. OpenAI will warn you that the free version of ChatGPT is “a free research preview.” For the Plus version, you’ll see an “upgrade to Plus” button on the left side of the home page.

For businesses, ChatGPT can also write and debug code, as well as create reports, presentations, emails and websites. In general, ChatGPT can draft the kind of prose you’d likely use for work (“Write an email accepting an invitation to speak at a cybersecurity conference.”). ChatGPT can answer questions (“What are similar books to [xyz]?”) as well. Microsoft showed off these features in its announcement that OpenAI is coming to Word and some other parts of the 365 business suite.

ChatGPT has historically not ‘remembered’ information from one conversation to another. However, starting on July 20, ChatGPT Plus members can use a feature called custom instructions, which is currently in beta, to make sure the AI remembers certain things about them. For example, it can remember a specific user tends to want content for a business audience, or, conversely, for third graders. Custom instructions will be rolled out to all users in the coming weeks, OpenAI said. It is not, however, available in the UK and EU.

ChatGPT app for iOS

On May 18, 2023, OpenAI announced the launch of the free ChatGPT app for iOS. The company stated the app syncs your history across devices, and that it integrates with its open-source speech-recognition system Whisper. On the iOS app, OpenAI said ChatGPT Plus subscribers get exclusive access to GPT-4’s capabilities, early access to features and faster response times.

OpenAI started this rollout in the U.S. As of May 24 it expanded to 11 more countries — Albania, Croatia, France, Germany, Ireland, Jamaica, Korea, New Zealand, Nicaragua, Nigeria, and the UK, with more expected to follow.

ChatGPT app for Android

ChatGPT for Android dropped on July 25, 2023 for users in the US, India, Bangladesh, and Brazil. Android users in those countries can get the app through the Google Play Store now. Additional countries will gain access over the next week, OpenAI said.

Browse with Bing

ChatGPT Plus subscribers had access to Browse with Bing, a feature in beta which enables ChatGPT to answer questions using accurate information pulled from the Bing search engine.

Browse with Bing was disabled on July 3, 2023 out of “an abundance of caution,” OpenAI wrote.

“We have learned that the ChatGPT Browse beta can occasionally display content in ways we don’t want,” said OpenAI scaling support lead Michael Schade in a help page. “For example, if a user specifically asks for a URL’s full text, it might inadvertently fulfill this request.”

OpenAI is working on re-establishing this feature, but has not announced a firm timeline for when it might be available again.

ChatGPT updates and OpenAI API news

OpenAI continues to update ChatGPT and its other services with developer-focused changes.

OpenAI’s bug bounty program

OpenAI started a bug bounty program on April 12, offering between $200 and $20,000 to ethical hackers who find vulnerabilities in the code. More critical vulnerabilities net larger bounties.

OpenAI isn’t looking for solutions to problems with ChatGPT’s content (e.g., the known “hallucinations”); instead, the organization wants hackers to report authentication issues, data exposure, payments issues, security issues with the plugin creation system and more. Details about the bug bounty program can be found on Bugcrowd.

Web browsing and plugins

GPTPlus users gained access to a beta version of web browsing and Plugins on the week of May 12. The beta includes web browsing mode, in which ChatGPT will sometimes access the internet to pull in information about current events.

Secondly, the beta version of ChatGPT will call on third-party plugins at the appropriate times if the user enables them. Third-party plugins can be accessed in the Plugin Store under Plugins in the model switcher. This opens ChatGPT up to more than 70 third-party plugins.

June 2023 API and pricing updates

On June 13, OpenAI added function calling to the Chat Completions API; reduced the price of their embeddings model (which helps the model interpret tokens); and reduced the price of input tokens for GPT-3.5 -turbo, one of the subscription models for the GPT 3.5 model.

With function calling, developers can describe functions to GPT-4 or GPT-3.5 turbo and the AI will return a JSON object which can call those functions. This could be used to create chatbot tools that call external plugins, convert natural language into database queries or API calls, or extract structured data from text.

Other announcements from OpenAI’s June 13 blog post include:

• updated and more steerable versions of GPT-4 and GPT-3.5-turbo
• new 16K context version of GPT-3.5-turbo (compared to the standard 4Kversion)
• Applications using GPT-3.5-turbo, GPT-4, and GPT-4-32K will automatically be upgraded to new models on June 27th.

Code interpreter

On July 6, OpenAI made ChatGPT’s code interpreter function available to all ChatGPT Plus users. The Code interpreter is an in-house plug-in with which ChatGPT can run code to analyze data, solve math problems, create charts, edit files, and similar tasks. It functions using a Python interpreter in a sandboxed, firewalled execution environment in a persistent session the length of the chat conversation, OpenAI said in their blog post.

Code interpreter is available in beta by taking the following steps in a ChatGPT Plus account:

Criticisms of generative AI like ChatGPT

With more and more organizations adopting generative AI, many questions arise. Will AI be able to fill jobs currently held by humans? What privacy and ethical concerns does it raise? These questions apply to both ChatGPT and its competitors, since any generative AI can perform similar tasks.

Will ChatGPT result in people losing jobs?

Whether ChatGPT will take jobs away from humans is impossible to predict. Goldman Sachs says in an April report that a quarter to a half of humans’ workloads could be automated with generative AI. The financial institution notes that doesn’t necessarily mean those jobs will disappear – instead, most will be “only partially exposed to automation” – and it may lead to up to a 7% increase in global GDP.

Roles that are repetitive or based on very specific rules are most likely to be able to be performed by AI, Steven Miller, professor emeritus of information systems at Singapore Management University, told CNBC.

ChatGPT could lead to new job roles being created, too. At the very least, people will be needed to prompt, train and audit AI like ChatGPT. Most likely, we’ll see the kind of shuffle that comes with any major technological shift as some jobs change and others do not.

Some experts refer to the current wave of AI as similar to the early days of the internet. Technological limitations still exist, and some estimations about how many jobs would be lost through automation have proven exaggerated in the past. The IEEE points out that the AI industry will need to be aware of hardware limitations and costs. Companies may not find it practical to spend enough money on AI services in order to replace a large percentage of their workforce. Paying users of ChatGPT can make a maximum of 25 GPT-4 queries every three hours, IEEE points out.

In some jobs, the AI may remove the need for a first draft, MIT labor economics professor David Autor said in an interview with CBS MoneyWatch. A human will need to tweak the output and provide in a unique angle or more varied wording, but ChatGPT could write the bare bones version of a speech or a blog post.

SEE: How ChatGPT could enhance jobs instead of replacing them.

Ethical and privacy concerns about ChatGPT

Perhaps inspired by science fiction about AI taking over the earth, some high-profile players in tech urge caution about giving AI too much free rein. On March 22, 2023, a petition and open letter signed by Elon Musk and many others urged companies to pause large AI development until more safeguards can be built in.

Ethics questions to ask when using generative AI

ChatGPT opens up questions about the ethics of using written content created by the algorithm. Posts created by AI should be clearly marked as such, but what about more casual communication such as emails? Business leaders should establish guidelines for when to be transparent about the use of ChatGPT or other AI at work.

OpenAI cautions that its products are not to be used for decisions in law enforcement or global politics. Privacy, which is perhaps a more pressing concern than global domination, led Italy to ban ChatGPT. OpenAI has since stated it wants to find a way to let ChatGPT work within the European Union’s strict privacy rules.

OpenAI’s new privacy update allows users to exclude themselves from training data

On April 25, 2023, OpenAI announced it has added a Chat History & Training setting that lets users turn off their ChatGPT chat history, preventing future versions of OpenAI’s large language models from training on those conversations. To find this option, click on your account name, which will display as your email address. Select Settings > Data Controls > Chat History & Training.

As of now, if this setting is not selected, user data will be fed back into the AI to train it on producing more naturalistic and useful responses.

OpenAI filters out personally identifiable information from the training data, OpenAI told Bloomberg. As of April 2023, users can get a copy of their ChatGPT chats and see what training data they have produced. OpenAI also plans to launch an enterprise subscription plan in which users’ data will not be shared by default.

Malicious uses of generative AI

Another potential problem comes from people using generative AI like ChatGPT to draft business email compromise messages or other attacks. Threat actors have created WormGPT, an application specifically for drafting malicious emails and customizing them to the prospective victims. Email security company SlashNext discovered WormGPT being used on black hat forums. WormGPT doesn’t actually share any genes with OpenAI’s ChatGPT; instead, the threat-oriented AI is based on GPT-J, a large language model from EleutherAI.

What are ChatGPT’s competitors?

ChatGPT’s primary competitors are or could be Google’s Bard, Baidu’s Ernie, DeepMind’s Sparrow and Meta’s BlenderBot.

Google’s Bard

ChatGPT’s main competitor is Bard, Google’s AI generative AI chatbot. People who would like to try Bard’s chat function need to join a waitlist.

Now Google plans to add Bard into search. In comparison to ChatGPT, Bard focuses more on creating prose that sounds like a human could have spoken it naturally and less on being able to answer any question. Bard is built on Google’s Language Model for Dialogue Applications.

While Microsoft is ahead of the pack right now in terms of providing chat functions to productivity software, the company lags behind in terms of its search engine Bing. Google decision-makers allegedly pivoted to urgently roll out a competitor for Microsoft’s decision to add generative AI to Bing search. (Meanwhile, ChatGPT helped Bing reach 100 million daily users.)

Baidu’s Ernie

The Chinese search engine Baidu plans to add a chatbot called Ernie. Baidu announced the upcoming change on March 16, at which point the initial showing disappointed investors.

DeepMind’s Sparrow

OpenAI competes with DeepMind, an artificial intelligence research laboratory owned by Alphabet. The two organizations are significantly different in terms of their aims. DeepMind focuses more on research and has not yet come out with a public-facing chatbot. DeepMind does have Sparrow, a chatbot designed specifically to help AI communicate in a way that is “helpful, correct and harmless.” DeepMind founder Demis Hassabis told The Independent in January 2023 that DeepMind may release a private beta version of Sparrow later in 2023.

Meta’s Llama 2 and BlenderBot

Meta released BlenderBot in August 2022. The prototype BlenderBot from the company behind Facebook focuses on being able to chat, providing short, conversational replies rather than full paragraphs.

Meta also has Llama 2, a foundational model competitive with the GPT-4 engine behind ChatGPT.

What about Apple?

According to The New York Times, Apple is working on leveraging the tech it has, especially Siri, to create a ChatGPT rival. More information about what the final product might look like is thin on the ground for now.

The future of AI in business

Will ChatGPT be common in online products in the future or is it a technological innovation forever in search of a greater use case? Today its “intelligence” is clearly still in the beginning stages, with OpenAI including disclaimers about inappropriate content or incorrect “hallucinations.” ChatGPT may put the words in a coherent order, but it won’t necessarily keep the facts straight.

GPT-3.5 and GPT-4 may also be getting worse at math. An August 2023 report from Stanford University and the University of California, Berkeley noted this “drift,” or gradual erosion of the ability to perform tasks like identifying prime numbers. Their theories as to why it’s happening include reduced ability to follow chain-of-thought (or, roughly, step-by-step) instructions.

In July 2023, two MIT economics graduate students conducted a study of 453 professionals. They found that people who used ChatGPT for writing tasks – such as producing press releases, short reports, or analysis plans – took 40% less time to finish their tasks than a control group that was not encouraged to use the generative AI. These professionals were then scored by their peers. On average, they received grades 18% higher than those in the control group (who did not use AI). This provides some qualitative data on the effect ChatGPT could have on white-collar work.

“Participants with weaker skills benefited the most from ChatGPT, which carries policy implications for efforts to reduce productivity inequality through AI,” wrote the authors of the study, Shakked Noy and Whitney Zhang.

Overall, Noy and Zhang maintained that widespread use of ChatGPT for writing tasks could have both positive and negative impacts in the workplace and the labor market.

Meanwhile, AI announcements that go viral can be good or bad news for investors. Microsoft’s stock price rose after the announcement of GPT-4, while Google’s stock dropped when Bard performed badly in a demonstration.

OpenAI saw visitor numbers to the ChatGPT website drop for the first time since its release in November 2022 in June 2023. According to Similarweb, worldwide unique visitors dropped 5.7% from May to June. Global desktop and mobile web traffic dropped 9.7%. ChatGPT still receives more worldwide visitors than Microsoft’s in-house AI at Bing.com. The shine may have worn off chat AI, although it’s too early to tell whether the business world will also start to cool on this trendy technology.

What’s next for OpenAI?

For now, OpenAI says it isn’t training GPT-5, the likely successor to today’s model. In a talk at MIT reported on by The Verge, OpenAI CEO Sam Altman pushed back against the open letter – an earlier draft of which had stated that a 5th generation was on the way; primarily, he criticized the letter’s lack of technical specificity.

“We are doing other things on top of GPT-4 that I think have all sorts of safety issues that are important to address and were totally left out of the letter,” Altman said.

He said no one should expect to see a GPT-5 rollout “for some time.”