Yes, AI Doomsday Scenario IS a Hoax.

It was upsetting to listen to today’s episode of The Daily. Simply put, it contained so much BS.

But thanks to the episode, I came to understand what Jacob Coxon’s job was at Anthropic and perhaps at OpenAI. He said, “I was part of the research team that would take the data that we scraped, and then decide which bits were good to train on.”

Essentially, he was part of a team that performed ETL (extract, transform, and load). ETL is a fancy term for getting, cleaning, and preparing data for actual analysis (an example of actual analysis is finding patterns in data, aka machine learning, which underlies modern AI). I bet his job involved manually looking through scraped data using some kind of computer program, identifying the less messy data, and cleaning up the messier data when necessary.

The point is that his skill set and capabilities have little to do with actual AI development. Simply put, he doesn’t know AI technically.

Another thing that reveals his limited knowledge of AI is his surprise at AI models solving complex math problems that humans struggle with. Well, solving a math problem, no matter how complicated it might be, is one of the things computers are particularly good at—as long as there is enough computing power. Thanks to continuously advancing chip technology (i.e., Moore’s Law), computers can now perform computations that would have seemed unimaginable just a few decades ago.

I also find the following post by an Anthropic employee (?) funny: “We really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade.”

Where is the 10% coming from? Why 10%? Is 10% dangerous? I’m just speechless.

Let me close by pointing out an AI developed in the ’60s. MIT professor Joseph Weizenbaum developed an experimental psychotherapy chatbot called ELIZA. The following is a quote from Wikipedia:

“Weizenbaum intended the program as a method to explore communication between humans and machines. He was surprised that some people, including his secretary, attributed human-like feelings to the computer program, a phenomenon that came to be called the ELIZA effect. ... While ELIZA was capable of engaging in discourse, it could not converse with true understanding. However, many early users were convinced of ELIZA’s intelligence and understanding, despite Weizenbaum’s insistence to the contrary.”

This was the ’60s—ELIZA was caveman-level technology. With all the human-like terms that current AI chatbots deploy (e.g., “thinking” instead of “processing”), no wonder people in 2026 think computers by themselves can go rogue.

The entire humanity could be wiped out by AI.

That's another level of fearmongering. Well, Jacob Coxon, a researcher who once worked at OpenAI and Anthropic, thinks that could happen by the end of the next decade. We may have a maximum of 15 years remaining on this planet.

Let me start with his credentials. I didn't search very hard. But he is young and was a researcher, perhaps a developer. Employees at that level simply tend to do what they are assigned to do, which is typically a very small portion of the entire technology being built. As a result, they don't get to form the big picture or fully understand the implications of the technology.

I watched a full interview Coxon did with Llamas. He seems to be passing along what may have been scuttlebutt at both companies. He didn't offer clear explanations for how AI will end humanity. His responses were generic: AI could infiltrate infrastructure, or AI could release some lethal viruses from controlled labs into the world.

Can AI go rogue? Absolutely. Remember, your simple computer right in front of you malfunctions every now and then. Also, AI designers and developers are human and cannot create an AI model that is completely free from errors.

Will AI have free will? Never.

There is nothing new in what Coxon is saying. He is just getting massive attention because he worked at both OpenAI and Anthropic.

Can AI malfunction and wipe us all off the planet? Maybe, because anything is possible. But before that happens, a crazy dude will nuke us off the planet first.

Chill out.

Which jobs will be replaced by AI?

Picking a major might be the most consequential decision one can make in life. In the current AI era, it is crucial to think through which jobs will be displaced by AI.

The good news is that the displacement process will be slow and that, in the end, there will still be demand for humans in heavily AI-affected job sectors. So, if you are passionate about a certain subject or major, you should go for it. Then, become one of the best in your area.

The bad news is that some jobs will be disproportionately affected by AI. To help you determine whether your job will fall into this category, let me offer you a succinct framework:

First, consider whether your job involves rule-based work. I know that even in rule-based work, exceptions exist, and AI may not be able to pick up on what may seem like complicated rules (at least to humans). That’s what many people in rule-based professions loudly claim. But AI will improve sooner rather than later. AI’s error rate will become much lower than that of the average human expert—if it isn’t already lower in many areas.

Second, consider whether there is a lot of digital data. Modern AI systems are typically based on machine learning, and the key to achieving accurate outcomes with machine learning is the quality and quantity of data. So, if a profession has generated a lot of digital data—or will do so in the future—consider other professions.

Third, consider legal and social issues. Even if a profession involves a lot of rule-based work or has accumulated a lot of digital data, it will not be easily replaced by AI if acquiring or using the necessary data is legally prohibited. But laws can change. So, the above two points matter more in your decision. In some cases, there will simply be social backlash against certain AI systems. But don’t count on it for your job security. There are many counterforces, too, such as corporate lobbyists.

The final point is this: consider whether replacing your job with AI is a commercially viable effort for a technology company. Even if your job can be easily replaced by AI, technology companies will not develop AI systems if your job is very niche or exists in a small market. In that case, your job may not be commercially attractive to AI companies. But again, don’t count on it. Your job may not be the main target but may become collateral damage.

The most important considerations, therefore, are whether much of your job is rule-based and whether your profession has built up a large amount of digital data—or is likely to do so in the future.

In Response to “What A.I. Is Actually Doing to the Economy”pt.2

Another thing that bothers me a lot about this episode can be essentially summarized in the following questions:

What should kids do? Will colleges still be necessary in this AI era?

This episode is not alone in raising those questions.But this IS PRECISELY the time to go to college.

You can even major in computer science and get a great job—as long as you focus on learning about and building AI.

Go to college, pick a major, pour everything you’ve got into it, and build domain knowledge and skills. Along the way, explore and use AI tools in a way that enhances your learning and professional skills. Become more mature and well-rounded. No future is more certain than another. To be prepared for uncertainty, do not overlook your liberal arts education (often called general education).

Colleges are not static—they, too, are adapting to AI to properly prepare students for the AI future.

Also remember, the knowledge economy won’t regress back to a labor-intensive economy. Labor-intensive jobs are the prime targets of AI and its associated robotics.

In Response to “What A.I. Is Actually Doing to the Economy”

I like The Daily. Its episodes are informative, polished, and clearly require substantial effort and resources to produce. But episodes like today’s also make me question the reliability of its coverage on subjects I know less about, such as law, medicine, or public policy.

One thing digital technology has done well for humanity is that it allows the public to hear directly from scholars who study technology. So, as a technology scholar, let me offer a different perspective.

The subtitle of the episode asks: “Some data suggest that artificial intelligence is causing job losses. Other sources show the opposite. Why is it so hard to figure out what’s going on?” The answer is not especially mysterious. Researchers—whether in academia, industry, or government—use different datasets, variables, analytical methods, assumptions, and time periods. These choices can produce different findings.

Equally important, researchers are human. We have biases, whether intentional or unintentional, and limited cognitive abilities. We do not know everything. We misremember things, overlook evidence, and sometimes see what we expect to see. Social science does not usually produce objective truth in the way many people imagine. It produces provisional explanations that become credible through evidence, criticism, replication, and, eventually, some degree of scholarly consensus. Disagreement is not necessarily evidence that research has failed. It is often how research works.

My larger concern with the episode, however, is its fearmongering tone. One of its central claims was that the Internet revolution of the 1990s unfolded gradually. Technology entered workplaces slowly enough for people to adjust. Layoffs happened over time, and some workers were even able to finish their careers before the full effects arrived.

The episode then avoided making a direct prediction about the AI era. But the implication was difficult to miss: this time will be different. AI will spread too quickly. Workers will not have enough time to adapt. Large-scale job losses may arrive suddenly.

Really? What is the evidence? Is it the rapid adoption of GenAI tools? News stories about AI-related layoffs? The apparent intelligence of systems such as ChatGPT and Claude? The speed at which these systems seem to be improving?

Not so fast. Much of what we are seeing is familiar. AI is surrounded by enormous hype, as many previous technologies were. Premature technologies are being aggressively promoted by vendors and amplified by the media, as they have been before. Many organizations are adopting so-called AI technologies partly because other organizations are doing so, as they have before.

Even thoughtful companies and capable employees need considerable time to determine how a new technology can actually create value. Decades of research in information systems (IS) and management suggest that technological change is rarely determined by technical capability alone. Organizations are constrained by politics, institutions, routines, regulations, incentives, skills, budgets, and ordinary human resistance.

Technology may advance quickly, but organizations usually do not. That distinction matters.

Yes, AI will contribute to layoffs. New technologies often do. Yes, some professions will be affected more severely than others. That is also not new. But those observations do not justify the claim that workers will have no time to adjust or that sudden, widespread firings are inevitable in the next few years.

I do not see convincing evidence that what people describe as AI is fundamentally different from previous waves of technological change.

Let us not mistake fear for insight.

Critical Thinking Is Overrated

I hear the importance of critical thinking invoked repeatedly—both within higher education and in broader public discourse. Over time, I grew weary of this and decided to engage in some “critical thinking” about the importance of critical thinking itself.

Let’s begin with the definition. Critical thinking means thinking critically about what is presented to you. Then, how does one develop this ability? In many cases, all it requires is a willingness to question what one encounters. It really is that simple.

Yet, college courses often bend over backward to demonstrate how they explicitly “incorporate critical thinking.” If I present a piece of information or knowledge to you, can you assess whether it is true or false? Just question it. The moment you do so, you are exercising your capacity for critical thinking. So, if we want to instill critical thinking in future leaders, our task is surprisingly simple: create environments in which students feel free to question anything.

But there is another, more demanding case where genuine critical thinking is actually required. If you truly want to determine whether something is true or false, questioning alone is not enough. You need domain knowledge. In other words, you need to learn.

Let’s help our future leaders learn deeply and independently.

Perhaps we should place less emphasis on the abstract, catch-all notion of “critical thinking” and more on cultivating independent learning. After all, that is often what people really mean when they invoke the C-word.

Painful Loss but Overall Good Season

My Oklahoma Sooners were eliminated by Alabama this past Friday in the first round of the College Football Playoff. Being a Sooner through and through, the game was hard to watch, and the loss was heartbreaking. I am still mildly depressed. But looking back, it was a great season overall. I am so proud of my Sooners. I am so much looking forward to 2026.

On Calling AI a “General Use Technology”

One of my favorite podcasts is The Journal by The Wall Street Journal. But in today’s episode—“China and the U.S. Are in a Race for AI Supremacy”—I heard something that genuinely me.

The reporter stated, “This [referring to AI] is the first general use technology we've seen come along since the internet, and so it affects potentially everything.”

Comments like this—coming from a reporter for a major newspaper—spread misunderstanding about AI and mislead a wide range of listeners. AI is not a “general use technology.” It is not one coherent thing. What we commonly call AI is a sprawling family of techniques, applications, and systems that vary dramatically in purpose, capability, and design.

The episode’s underlying narrative is that the U.S. and China are locked in a race to build “artificial general intelligence” (AGI). But is AGI a general-use technology? Only if we define AGI as a system that “knows everything deeply.” Under that definition, a hypothetical future ChatGPT could be called AGI. But such a technology is neither necessary nor even desirable. In many fields—medicine, law, finance—specialized systems already surpass (or will soon surpass) human expertise in narrow domains. What good reason do we have to combine all such domain-specific “intelligent” systems into a single, unified AGI? None.

And are those domain-specific intelligent systems “general use technologies”? Again, no. They are built to solve specific problems, operate under specific constraints, and serve specific communities of practice.

So here is my plea: When we talk about AI, we must define—at least conceptually—the specific technology at hand. You don’t need to list a concrete product or application, but you do need to specify what kind of system you’re referring to. Otherwise, we are speaking in vague abstractions that obscure far more than they reveal.

Precision matters. Without it, public discourse on AI will continue to drift toward confusion—rather than understanding.

Teacher, Entertainer, Babysitter: The Hidden Triple Threat in Academia

In a recent episode of This IS Research Podcast, one of the co-hosts argued that a “superstar” professor in any field should not be teaching freshman classes. According to him, teaching first-year students requires nothing more than a blend of teacher, entertainer, and babysitter—hardly a good use of a superstar’s time.

Perhaps society does benefit when its academic superstars devote most of their energy to research. Even if we grant that possibility, isn’t that supposedly “simple” blend—a teacher, an entertainer, and a babysitter—actually a remarkable combination? I have rarely met individuals who genuinely possess this mix of skill, charisma, and patience.

If you are one of them, you bring tremendous value—especially now, as colleges confront the so-called demographic cliff. Your work with first-year students is not trivial; it is indispensable.

Why “Different” Is the Default, Not the Insight

I recently read “ Agentic AI at Scale: Redefining Management for a Superhuman Workforce ” in MIT Sloan Management Review. The article notes that nearly 70% of surveyed experts claim that agentic AI accountability demands entirely new management approaches, while 25% disagree. I applaud the minority.

It is remarkably easy to label every new invention as fundamentally different. We do it because:

1. Novelty is socially inherited, not independently concluded.
   Experts and observers often believe a technology is new because they heard others say it is. Repetition manufactures inevitability.
2. Agreement is cognitively and reputationally cheap.
   Conforming to the majority is easier than resisting it.
3. Continuity requires proof; difference requires none.
   Declaring “difference”—the popular choice—invites little challenge. Declaring continuity flips the burden: you must defend it. In a low-attention, high-velocity world, most avoid the cost.

Difference is often not insight—it’s echo, convenience, and safety in disguise. That alone makes it worth interrogating.

Doing Slow Work in a Fast Digital World

Much of my research looks at very fast phenomena—tweets, platform launches, market reactions. Yet the work itself is painfully slow. Data collection, coding, theorizing, and revising a paper over many years can feel out of sync with the speed of digital life.

I have come to see this tension as a feature, not a bug. Slowness gives us room to notice patterns that are invisible in the moment and to question “obvious” narratives. It is one way academia can add value in a world flooded with instant commentary.

First-Generation Paths and Paying It Forward

As a first-generation college graduate, I still remember how opaque universities felt when I was a student. Many unwritten rules were confusing, and chances to “get involved” or “network” did not feel designed with students like me in mind.

That memory shapes how I advise and teach. I try to make expectations explicit, open doors to research and projects, and remind students that their background is not a deficit but a source of insight. Education works best when it makes more future paths visible, not fewer.