An AI pioneer says the technology is 'limited' and won't replace humans anytime soon

[BumRushDaShow]

Source: NBC News

Dec. 27, 2025, 6:00 AM EST


NEW YORK — When Andrew Ng talks about AI, people listen — in classrooms, boardrooms and Silicon Valley. The researcher-turned-educator-turned-investor has become an AI statesman of sorts, co-founding Google Brain, which became part of Google’s flagship DeepMind division that now produces some of the world’s best AI systems, and serving as Chief Scientist of Chinese tech titan Baidu.

In today’s influencer-obsessed information landscape, Ng’s biggest claim to fame might be his credential as a “Top Voice” on LinkedIn, an honor the platform gives to a select few handpicked experts, with over 2.3 million followers.

Armed with decades of AI experience, Ng says he remains clear-eyed about AI’s abilities. “The tricky thing about AI is that it is amazing and it is also highly limited,” Ng told NBC News in an interview on the sidelines of his AI Developers Conference in November. “And understanding that balance of how amazing and how limited it is, that’s difficult.”

Over the past few years, generative AI has attracted hundreds of billions of dollars in investment as nearly every major tech company has pivoted towards the industry’s hottest topic. But in the last several months, many have questioned whether the surging investment has created a bubble now at risk of bursting due to persistent issues like hallucinations, AI’s involvement in mental health crises and increased regulatory scrutiny.

Read more: https://www.nbcnews.com/tech/innovation/andrew-ng-says-ai-limited-wont-replace-humans-anytime-soon-rcna246074

---

It IS a bubble and the amount of resources being expended to create datacenters that will not have enough power to operate nor enough staff knowledgeable to maintain the hardware itself (with all the focus on software programmers), looks like a catastrophe in the making.

[bucolic_frolic]

1. If it don't make pizza, it ain't worth nothing. /nt

[flashman13]

2. Might I add that you can only divide the pie of potential users so far. There is no way all of these data

center builders can gain enough users to pay any sort of reasonable return on investment. That's essentially why the bubble is going to burst.

[cstanleytech]

4. In the end I think what you'll see is Amazo, Microsoft and Google be the dominant three.

[flashman13]

5. In the end the big guys will gobble up everyone else for pennies on the dollar.

A vast number of investors are going to take it in the shorts.

[cstanleytech]

3. Well I got agree there as when it comes to writing they are extremely limited.

For example if you try one for say writing a story or something it'll initially be okay but as precedes it starts to lose track of character relationships, past events and even the storyline.

[not fooled]

7. Yep.

AI seems to me to be useful for answering simple, factual questions that benefit from sweeping reviews of large amounts of information, Or, of course, performing well-defined rote tasks that can be automated.

But compare with a human brain when it comes to extrapolating from existing information to create something new, such as writing a story? No way. Or, in my experience, analyzing and interpreting complex visual images, such as of art and antiques. I use reverse image searches to price items I'm considering buying at auction. Google Images which I assume is AI based does a wretched job most of the time, frequently missing the era entirely or wrongly guessing the nature and function of an item. I'm not technically sophisticated enough to define why AI sucks at this, but...it does.

[cstanleytech]

14. Even it's factual questions can be flawed so you should always verify as some Trump lawyers are learning right now.

[highplainsdem]

11. So don't use AI for writing. Anything you write with it isn't your work anyway and can't be copyrighted.

[cstanleytech]

13. I don't, I have tested it out though and it's just not at the point where it'll replace a human being.

[William Seger]

6. One serious limit: AI bots are completely incapable of actual logic

When AI bots imitate human speech, they might sound perfectly logical, but that's really just accidental and definitely not guaranteed. Actual logic requires evaluating the soundness of premises and the logical validity of conclusions (in the sense that they cannot be false if the premises are true), NEITHER of which AI bots are capable of.

[Shipwack]

8. While he's correct, it doesn't matter.

What matters is whether the pointy hair bosses believe this. A lot of them don’t, or have been sold a bill of goods by the consultants they overpayed to advise them.

The “smartest” of them know AI is crappy, but they feel it’s “good enough” to replace a whole lot of people. A whole slew of entry level writing jobs at publications have been taken over by bots. They haven’t thought about where new mid level writers are going to come from… Then again, maybe they’re hoping that by the time that’s an issue AI will have advanced enough.

[ToxMarz]

9. AI isn't really that much intelligence (for now at least), it is automation on steroids

it requires a huge amount of proccessing power to achieve, but it is basically the capablities we've already had with the much greater proccessing power through high end (expensive) chips and gpu's locally and mutli--billion $$ cloud data centers. It is mostly running routines and instructions we've had previously, but with a better more human like interface and expanded features that were considered a waste of the capped computing power generally available previously. As such more trained personel were previously needed to interface with the automation as it wasn't so user friendly.

[mdbl]

10. I heard the same thing from an industry insider

He told me that it will only be able to do menial focused tasks. If AI is completely relied upon, it will give poor customer service due to it's limitations. When I asked why there is so much hype around it he said it was because the entire AI stock market bubble is reliant upon convincing investors that it will save all kinds of money and be sold to everyone. He informed me that companies that are buying and/or considering it are already seeing it's limitations and aren't happy with it many of the implementations. As soon as the market gets that, the bubble will burst. The sad part is, they will be forcing down everyone's throats for the time being and we as the general public just have to deal with shitty service.

[LudwigPastorius]

12. The venture capital bubble may burst, but that's not going to stop the research.

Governments, including the U.S. government, will take up the slack in any funding because they are convinced that achieving Artificial General Intelligence (if it's possible) is a matter of national security.

And, while generative AI and LLMs are getting all of the hype, there are other types of machine intelligence that are being pursued. World models, action models, multimodal modals, synaptic brain mapping and modeling, expert models, and reinforcement learning are all being developed. My guess is that if AGI is ever achieved, it will be a result of combining many approaches.

[highplainsdem]

15. That article is nothing but pro-AI hype from someone incapable of being objective about AI. He's

anti-regulation, so he probably backs Trump.

All those words about AI from him, and nothing about the AI companies having stolen the world's intellectual property.

He can babble about transparency by AI companies supposedly being better than regulation, but that"s bullshit and misdirection because the AI companies are opposed to transparency as well, and there's no way he isn't aware of that. They don't want to reveal their training data because they stole it.

He trivializes the harm done by chatbots, pretends to feel concern for the tragic deaths, but what he says next makes it clear chatbots getting people to commit suicide isn't anything he's really concerned about - at least not enough for him to think it's an adequate reason to regulate AI companies.

The paragraph below shows where his real focus is. Profits.

Ng said the first steps of creating AI models, referred to as the “training” or “pretraining” stages, “is where a lot of the questions are, where the very real questions are. When will the payoff for all of the capital expenses going into this training, when will they pay off?”

That training stage is where the theft of intellectual property starts. He says that's "where the very real questions are" - but it's obvious that to him the only real question is when companies can profit from that theft. Not if they should stop stealing and pay for what they already stole.

He's just another damn robber baron.