News TSMC execs allegedly dismissed OpenAI CEO Sam Altman as ‘podcasting bro’

Toggle sidebar Toggle sidebar
If TSMC isn't willing to work with Sam Altman, I would think Intel might have the capacity, maybe not the capability to match TSMC, but sometimes your best ability is your availability...
 
I really don't think "AI makers" will be the winners in this round of technology. It will be those who best leverage "AI"/ML who end up getting the "win" from "AI". Unless OpenAI and other can come up with something truly proprietary and not easily reproducible using open sourced software, it's going to be a long road IMO.
 
JamesJones44 said:
I really don't think "AI makers" will be the winners in this round of technology. It will be those who best leverage "AI"/ML who end up getting the "win" from "AI". Unless OpenAI and other can come up with something truly proprietary and not easily reproducible using open sourced software, it's going to be a long road IMO.
Click to expand...
Like the gold rush it's the people selling the pots and shovels that are raking in the money.

I think open AI won't be long for this world as meta and google get up to steam and microsoft starts doing their own thing
 
  • Like
Reactions: snemarch and JamesJones44
Karismatic mania can get you a long way but it doesn't work at all times.

TSMC execs know that the road forward is unlikely to be perfectly straight and so exponential acceleration may not be a sound strategy.

Applying some Eastern wisdom is probably prudent. Yin and Yang, even Wall Street finance bros know that the Bear comes with the Bull and with a boom comes a bust.
 
nobody is gonna invest that much $ into ai as there is no guarantee of profit.

"ai" is niche for most of the people alive (same as crypto but that had profit market)

even if you spend 10 yrs and unlimited funding there is no guarantee that ai will ever advance enough & be profitable for people to WANT to have it.
 
You want to remember that at least up to a few years ago the whole credit for the success (such as it is) of the LLM's has been scale, scale, scale. And the simplest projection is a linear (or better) increase with no limits. So if it's true, the first guy to a trillion dollars for scale, wins. Rules the world, perhaps.

This has been a question about AI since the early days - "maybe there's some critical mass, like with uranium and plutonium, and that's all there is to it!"

And the odds are that this is at least partially right, and partially wrong. Maybe scale still has a ways to go, but maybe the *cost* of scale can come down. Maybe some algorithm tuning can make training 10x, 100x, 100000x more efficient.

The human brain runs on about 20 watts and fits in goldfish bowl, so if we were just doing it right, wouldn't AI?

So my point? Oh yeah. The efficiency and cost of training is already coming down by a factor of 10x or so, and promises to continue, and Altman knows that (as much as he knows anything). This is just his act, his schtick, make of it what you will. OpenAI is doing some good work while he zooms around saying silly things.
 
Kondamin said:
It’s nicer than what I would have called someone that came to me asking for 36 fabs at a time when the competition is in trouble for trying to build a couple
Click to expand...
It's still weird, sounds like something a non-native speaker would think is a witty insult. If we are talking about not being nice, TSMC execs deserve to be called much worse things after they showed their true colours when dealing with AZ workers.
 
Heat_Fan89 said:
An interesting take from this guy and I have to ask similar questions as to why?

peakprosperity.com

What Problem is the Data Center Frenzy Aimed at? |

I think this is big. I need your help unraveling the true underlying reasons for the completely out-of-proportion data center/AI build out. What's driving it, and how should we prepare? I present my case here as a starting point; let's figure this out together.
peakprosperity.com peakprosperity.com
Click to expand...
It's a combination of things. There's been an increasing build-out of data centers since about twenty years ago with Yahoo and Google and then Facebook, AWS, Azure, and a few others, all the hyperscale guys. And now with all the ESG raising energy prices and refusing to build reliable power plants, even modest continued growth is a problem. And this modular nuke solution has been around for decades just looking to get off the ground. It's a match made in heaven.

The AI training task has been an energy pig now for three or four years, and they believe, it's part of their catechism, that more is better, and moar, AND MORE, scale, scale, scale. Like bitcoin only moreso. And as everybody and their monkey wants to get into the act, we'll need more more more power. Now none of this is really workable but that's OK because almost none of it is true. But if someone wants to build out a dozen or a hundred modular nukes, this is probably an excellent time to do so.
 
JRStern said:
It's a combination of things. There's been an increasing build-out of data centers since about twenty years ago with Yahoo and Google and then Facebook, AWS, Azure, and a few others, all the hyperscale guys. And now with all the ESG raising energy prices and refusing to build reliable power plants, even modest continued growth is a problem. And this modular nuke solution has been around for decades just looking to get off the ground. It's a match made in heaven.

The AI training task has been an energy pig now for three or four years, and they believe, it's part of their catechism, that more is better, and moar, AND MORE, scale, scale, scale. Like bitcoin only moreso. And as everybody and their monkey wants to get into the act, we'll need more more more power. Now none of this is really workable but that's OK because almost none of it is true. But if someone wants to build out a dozen or a hundred modular nukes, this is probably an excellent time to do so.
Click to expand...

If I have an excellent plan to build multiple clean efficient power plants producing the electricity I know society will need and be happy to pay for in the future I'm willing to throw out some "AI marketing" to get the financing and approvals.

A dozen working modular nuclear power plants will have value.
 
  • Like
Reactions: JRStern
Elusive Ruse said:
It's still weird, sounds like something a non-native speaker would think is a witty insult. If we are talking about not being nice, TSMC execs deserve to be called much worse things after they showed their true colours when dealing with AZ workers.
Click to expand...
That’s because it’s a pretty witty insult, without calling him a charlatan who drinks his own poison that says nice things that sound plausible on the surface but are totally bunk to specialists on the subject.

He can go sit on the Joe rogan show with Elon musk talking out their ass about building a solar powered hyperloop to mars where they will build ai data centres that don’t need special cooling because it’s cold on mars
 
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom