Satya Nadella warns AI must go mainstream to avoid becoming a bubble

[Alfonso Maruccia]

Editor's take: The cost of consumer hardware is skyrocketing and people are losing their jobs, yet Satya Nadella wants even more AI services and products to reach a wider audience. The technology has the potential to drive growth, but it must go beyond the IT sector to "touch" every aspect of every industry.

While remaining an enthusiastic proponent of AI, Satya Nadella now warns that the technology could eventually face the consequences of a massively speculative bubble. The Microsoft CEO spoke on the first day of the World Economic Forum in Davos, explaining how large language models, chatbots, and other recent AI innovations can avoid the fate of the dotcom crash.

Nadella believes AI may succeed only if it achieves widespread adoption beyond the technology sector. Chatbots and LLMs need to be integrated into multiple industries – and even reach countries beyond the wealthiest economies – to justify the unprecedented capital expenditures fueling their development.

"For this not to be a bubble by definition, it requires that the benefits of this are much more evenly spread," Nadella said in a talk with BlackRock chief Larry Fink.

Limited adoption by Big Tech alone, Nadella warned, would be a clear sign that AI is indeed a financial bubble. Yet he remains optimistic that AI has the potential to deliver truly transformative changes across industries, starting with pharmaceuticals and the development of new drugs.

Nadella leaned on his best corporate lingo to impress WEF attendees. According to the Microsoft CEO, AI is set to build on the advances the IT industry brought with cloud and mobile solutions. The technology, he claims, will spread rapidly, boost productivity in a significant way, and generally expand the global economy, both in developed nations and beyond.

He also suggested that companies could eventually resolve the intellectual property and trademark challenges arising from the questionable training practices used by AI developers. Nadella recommends that enterprise organizations "distill" from larger models, retraining LLMs on their own data and specific business contexts. This approach, he argues, will allow industries to leverage multiple models, including open-source ones, without depending on a single AI provider.

The Davos talks confirmed that Nadella remains a true believer in AI's transformative potential, but they also revealed more about the state of the industry than its future. Many business leaders are still unclear on how AI can be genuinely useful, while the Microsoft CEO is determined to ensure that disgruntled Windows users no longer dismiss AI as slop.

Permalink to story:

Satya Nadella warns AI must go mainstream to avoid becoming a bubble

 

[Theinsanegamer]

So the liar's facade finally cracks.

If you need to push something on consumers to avoid it "being a bubble", then it is a bubble, and you are looking for marks to keep your scam coming.

Nothing about current "AI" tools justifies even a mere fraction of their current valuation and the cash burn is utterly unsustainable. Using money you dont have, to buy RAM that doesnt exist, from fabs that dont currently exist, for GPUs you havent made, to put in datacenters that havent been built, for a product that loses money is serious dotcom bubble energy.

 

[PXAbstraction]

Hate to tell ya' Satya...

 

[winjer]

Snake oil salesman trying to convince more people to buy his slop.
Thanks, but no thanks.

 

[jbright]

There’s a quiet irony in talking about spreading AI benefits globally while consumer hardware prices and cloud costs keep rising. Adoption is one thing, affordability and usefulness are another.

 

[GodisanAtheist]

Time to put a little pressure on ole Donald by taking a bit of air outta the market.

They'll get some backroom deals from the WH and then it will be back to "full steam ahead" messaging.

 

[brunocasarini]

In another words: "please, help us recover the billions we poured into AI before the bubble bursts".

 

[passwordistaco]

The bubble cannot burst soon enough.

 

[bviktor]

DD remind me again why I'm supposed to care

 

[OortCloud]

LLMs have pretty much hit a brick wall now. They have basically thrown all the data on the entire internet (often illegally) into these things and it turns out they are still super unreliable. It's barn-door engineering, just hurling vast quantities of text and images into Neural Network's with very little understanding of how those NN's really process this and hoping to get something sensible out the other end. It's not truly intelligent in any way, it's a very inefficient and expensive parlour trick.
Somebody far brighter than me explaining this:
https://www.mindprison.cc/p/no-progress-toward-agi-llm-braindead-unreliable

 

[Beerfloat]

AI not becoming a bubble is not a purpose and should not be a goal. Having a technology and then trying to figure out what to do with it is the essence of putting the cart before the horse. If there was demand for a larger role for AI then this would not be a concern.

Instead, companies like Microsoft are hindering AI growth by pushing it in places where it's not particularly useful. Copilot now pops up in Windows, Office applications, on Microsoft websites, on Android phones and who knows where next. The way they are going about it induces resentment and eye-rolling more than engagement.

 

[ZedRM]

@ Satya Nadella

You are out of touch with the general public. No one wants AI to be mainstream and they don't want it on their PC's. It doesn't serve us and it's not trustworthy. Take a hint Mr Gump!

Pop the freakin bubble so life, and prices, can get back to normal!

 

[ScottSoapbox]

The reverse is also true.

AI wouldn’t need to impact every facet of every industry if they weren’t spending money to build it like that impact was an inevitability.

 

[IMABEARLOL]

Hey Microsoft, make Windows, make Office, make copilot if you want. But keep them separate and almighty capitalism will determine the product that people want.

Stop resorting to forcefeeding your consumer a product they clearly do not want to justify your sunk cost fallacy. A product with merit will stand on it's own. You are ruining the products people do want.

 

[LimyG]

AI not becoming a bubble is not a purpose and should not be a goal. Having a technology and then trying to figure out what to do with it is the essence of putting the cart before the horse. If there was demand for a larger role for AI then this would not be a concern.

Instead, companies like Microsoft are hindering AI growth by pushing it in places where it's not particularly useful. Copilot now pops up in Windows, Office applications, on Microsoft websites, on Android phones and who knows where next. The way they are going about it induces resentment and eye-rolling more than engagement.

What you said, in particular the second paragraph about pushing it in places where it's not particularly useful is exactly how I feel.
I would add to that by saying, not wanted in the first place by many.

There is no denying that A.I. in the right hands by the right people, for instance medical research and a lot more is already valuable and showing real world benefits. -
- Breast cancer scans are one example. A.I. is better at spotting very early cancers that specialists often miss. There was a BBC program segment about this. Fantastic stuff!!

But putting all this Co-Pilot and more, "A.I." with plans to go "agentic," on everyones W11 OS by default is beyond the... I'll just say very annoying.

Finally, his way of attempting to offset a potential bubble, his wording, his snake oil like persona. Is .....

I can't say what I really think about this Sataya as it would rightfully be deleted by the Forum bosses.

 

[LimyG]

Hey Microsoft, make Windows, make Office, make copilot if you want. But keep them separate and almighty capitalism will determine the product that people want.

Stop resorting to forcefeeding your consumer a product they clearly do not want to justify your sunk cost fallacy. A product with merit will stand on it's own. You are ruining the products people do want.

Yep. 100% agree.

 

[Taroo]

Or what about no ? Let it be a bubble and let I finally explode.
This is the payback.

 

[Revolution 11]

"To avoid drowning, you must go over the Niagara Falls in a barrel" - Satya the Wise Prophet

How much of an ***** does he think we are?

 

[Later]

He can take all Ai and push it up somewhere the sun does not shine, and leave us alone, if we need Ai, will get some installed, if not, NO.

 

[LemmingO]

"If it swims like a duck, walks like a duck, and quacks like a duck..."

AI is a bubble. It will be 2007 all over again. Don't let your moronic leaders bail out any company that fed into this.

 

[BobHome]

So the liar's facade finally cracks.

If you need to push something on consumers to avoid it "being a bubble", then it is a bubble, and you are looking for marks to keep your scam coming.

Nothing about current "AI" tools justifies even a mere fraction of their current valuation and the cash burn is utterly unsustainable. Using money you dont have, to buy RAM that doesnt exist, from fabs that dont currently exist, for GPUs you havent made, to put in datacenters that havent been built, for a product that loses money is serious dotcom bubble energy.

His talk to "impress WEF attendees" made me think of Gates telling the Pentagon how great MS-DOS was supposed to be.

And... . . . . Pop!

 

[BobHome]

LLMs have pretty much hit a brick wall now. They have basically thrown all the data on the entire internet (often illegally) into these things and it turns out they are still super unreliable. It's barn-door engineering, just hurling vast quantities of text and images into Neural Network's with very little understanding of how those NN's really process this and hoping to get something sensible out the other end. It's not truly intelligent in any way, it's a very inefficient and expensive parlour trick.
Somebody far brighter than me explaining this:
https://www.mindprison.cc/p/no-progress-toward-agi-llm-braindead-unreliable

Well, since they gobbled up all the available info unabashedly and 25% - 46% is false--depending on which reports one reads--it's very easy for AI to hallucinate.

So, who is going to take the time to do it right???

 

[scavengerspc]

From what I have seen and heard, which I'm sure is a lot less than most here,
AI is already a bubble, or well on its way.

 

[amghwk]

It's already a bubble.