News Elon Musk Confirms 'TruthGPT' AI Project, Will Use Lots of GPUs

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.

TJ Hooker

Titan
Ambassador
I'm betting the name being close to Trump's "Truth Social" is far from a coincidence. You can bet this AI, if it actually gets off the ground, is going to be full of so-called "alternative facts"
Yeah, the choice of name on its own could have just been a coincidence. But when combined with the fact that Musk made this announcement on Tucker Carlson, along with Musk's general behavior over the last while, I can't help but have similar suspicions.
 
Last edited:

atomicWAR

Glorious
Ambassador
Elon seems to have some posters on here big mad.
Ah man have a sense a humor. 99% of what I said was in jest as what he proposes is a near impossible task for a variety of reasons like bias of programmers, mistakes in code, information used, programming models, etc. So much can go wrong so easily. And even if he does succeed whether its Musk, his board or humanity as a whole but we might not like what TruthGPT tells us...true or otherwise. And again who decides what is true and not? Its really sketchy across the board. And thats before you get into the possibility of evil AI. Think of what conflict an AI could wrought by controlling, altering or abridging the 'truth' to various groups.

AI is scary stuff I was doing my best to make light of. I find a good laugh leads to better personal health. BUT I can get where you are coming from all the same. When you try to make proper humor you always risk poking the bear, or you didn't try very hard in the first place. Regardless I think objectively speaking I just think its a bad idea. Which I find funny Musk would get behind something like this considering how he has screamed from the mountain tops the evils of AI. He even went as far calling AI the 'Immortal Dictator'. Seems like a bit of a 180 don't you think?

 
Ah man have a sense a humor. 99% of what I said was in jest as what he proposes is a near impossible task for a variety of reasons like bias of programmers, mistakes in code, information used, programming models, etc. So much can go wrong so easily. And even if he does succeed whether its Musk, his board or humanity as a whole but we might not like what TruthGPT tells us...true or otherwise. And again who decides what is true and not? Its really sketchy across the board. And thats before you get into the possibility of evil AI. Think of what conflict an AI could wrought by controlling, altering or abridging the 'truth' to various groups.

AI is scary stuff I was doing my best to make light of. I find a good laugh leads to better personal health. BUT I can get where you are coming from all the same. When you try to make proper humor you always risk poking the bear, or you didn't try very hard in the first place. Regardless I think objectively speaking I just think its a bad idea. Which I find funny Musk would get behind something like this considering how he has screamed from the mountain tops the evils of AI. He even went as far calling AI the 'Immortal Dictator'. Seems like a bit of a 180 don't you think?

Like I already posted on here ... I don't want to see AI involved in social media. I don't know what Elon's end game is but after seeing what he's done with Twitter since taking it over I have expectations that he means good with this latest endeavor.
 
  • Like
Reactions: Icepilot

atomicWAR

Glorious
Ambassador
Like I already posted on here ... I don't want to see AI involved in social media. I don't know what Elon's end game is but after seeing what he's done with Twitter since taking it over I have expectations that he means good with this latest endeavor.
Ok i get that but do you not think Musk is pulling a bit of a 180 on AI as of late? (honestly curious)
 
  • Like
Reactions: daworstplaya

bit_user

Polypheme
Ambassador
Yeah I can kind of see it now. Musk announces he succeeded in creating an AI for truthGPT on stage at some tech-con (pun intended)...

"Hey TruthGPT tell me a little about the CEO of Tesla and the owner of Twitter?", Musk says triumphantly.

"Elon Musk is the CEO of Tesla and owner of Twitter also known as the chief twit. He also makes space ships, gave his most recent child a weird name and is generally a Tool most of the tech community is to embarrassed to claim as one of their own." responds TruthGPT.
"Mirror, Mirror on the wall..."
 
  • Like
Reactions: atomicWAR

bit_user

Polypheme
Ambassador
I also get frustrated with how PC all the other AIs are. You know they could offer opinions, but they are programmed not to.
They're not constrained from offering opinions. They're constrained on controversial topics. People think this is about some political agenda, but really it's just about corporations wanting to avoid bad publicity.

What it says is a function of what it was trained on, not necessarily what's true. Language models are trained to learn what a certain type of speaker would say, or how to say things in a certain style. They shouldn't be seen as an oracle of truth.
 
They're not constrained from offering opinions. They're constrained on controversial topics. People think this is about some political agenda, but really it's just about corporations wanting to avoid bad publicity.

What it says is a function of what it was trained on, not necessarily what's true. Language models are trained to learn what a certain type of speaker would say, or how to say things in a certain style. They shouldn't be seen as an oracle of truth.
Says who?
 
The economics of running a social media network will demand it be AI-moderated. Especially in the era of AI-powered bots.
Musk has already reduced Twitter's workforce by 80%. Not that I have a problem with tech employees getting laid off. I feel just as bad for them losing their job as they did when me and thousands like myself were laid off from our jobs in the oil, gas & mining industries.
 

bit_user

Polypheme
Ambassador
So much can go wrong so easily. And even if he does succeed whether its Musk, his board or humanity as a whole but we might not like what TruthGPT tells us...true or otherwise. And again who decides what is true and not?
It could turn out to be the ultimate tool for debunking conspiracy theories, except that conspiracy theorists won't trust it. In the worst case, it could be a fountain of new or improved conspiracy theories.

We all know how sensitive AI models are to what you feed into them. As the newspaper industry continues to implode, there are fewer and fewer journalists doing investigation, gathering facts, and writing original journalism. Without a solid base of facts, it could end up relying on spin, false reports, and self-generated facts (referred to as "hallucinations").

thats before you get into the possibility of evil AI. Think of what conflict an AI could wrought by controlling, altering or abridging the 'truth' to various groups.
I worry more about faulty or biased AI, but also what a powerful tool AI bots can be for bad actors intentionally spreading disinformation and false narratives.
 

ngilbert

Prominent
Mar 30, 2023
6
7
515
I find this incredibly funny. Musk wants to make an algorithm that will evaluate the truth of a statement. The Turing machine, the spiritual predecessor of all of computing, mathematically proved that there will be logical statements that cannot be proven true or false in finite time.

In other words, Musk wants to do something that is proven impossible, using the successor of the thing that proved it to be impossible...
 

RedBear87

Commendable
Dec 1, 2021
153
120
1,760
I find this incredibly funny. Musk wants to make an algorithm that will evaluate the truth of a statement. The Turing machine, the spiritual predecessor of all of computing, mathematically proved that there will be logical statements that cannot be proven true or false in finite time.
I find it fascinating instead, by absurd it might very well turn out to be the best way to prove how all those charlatans posing as "fact checkers" are just human-powered propaganda bots...
 
  • Like
Reactions: Nathkrul and Why_Me

sam buddy

Distinguished
Nov 29, 2011
45
20
18,535
different in most people accept truth when solid backing of claim not just group mentality on something. (why most people can change their views when given evidence to counter their former understanding of something)
Change both instances of "most" with "some", and I totally agree :) (yep, lost faith in humanity a long time ago).

Musk's truth is far from real truth as we have seen during & leading up to his twitter ownership.
I see what you're saying, but he's not alone, everyone and their dog(e) is doing it, that's what I was saying in my first reply. Anyway, we'll have to wait and see how it develops, if ever.
 
Last edited:
  • Like
Reactions: bit_user

baboma

Prominent
Nov 3, 2022
200
182
760
I see several issues facing Musk's foray into AI, even without considering his prickly personality and capricious handling of Twitter.

He's well behind the frontrunners, Microsoft/OpenAI and Google. He's buying Nvidia GPUs and trying to recruit new personnel to staff up, while they're already making their own custom AI-accelerators and have established staffing and AI models. Both are going full steam ahead, locked in competition with each other. It's to see how a copycat will catch up, let alone surpass them.

Given his mistreatment of Twitter's employees and his strident right-wing ideology that alienates many people, it will be difficult attracting top-end talent. Sure, money talks, and reportedly he recruited some dude from DeepMind, but he'll need a lot more than just a handful of AI specialists. Tech workers by and large have a liberal bent, and AI techs are in high demand now. So the value proposition of jumping ship to work for a right-wing troll who has a track record of employee mistreatment, is patently unattractive. Just to wit Trump's Truth Social's recruitment issues (where Musk got his TruthGPT riff) and multiply that times 100.

His track record with AI development is poor, if Tesla's ongoing FSD debacle is any indication. The fact that he's calling his venture [XYZ]GPT means he's content to follow the frontrunners' footsteps with a GPT model, along with every other company. He'll be just one more copycat in a crowd of copycats, with minimal prospect of standing out. His explanation of making AI safer by "trying to understand the truth of the universe" is laughably ridiculous and smacks of a snake-oil pitch.

Musk's newly-formed X.AI company is yet another distraction. Already, his core company Tesla (which constitutes the core of his wealth) is suffering from lack of leadership. Its first-mover advantage is quickly eroding along with its stock price. No new products along with repeated price cuts don't portend a bright future for that company.

Let's be clear, Musk has had some outstanding achievements in service of society. SpaceX has advanced spaceflight with cheap reusable rockets, and has re-energized the US space efforts. Starlink likewise has advanced sat-Internet connectivity, and is the single most important contribution in the Ukraine-Russia war. Tesla has legitimized the notion of EVs and accelerated its adoption by the car industry.

These were all monumental accomplishments, and Musk deserves credit. But past performance does not predict future results. His latest attempt to remake Twitter was a disaster, and exposed his many failings. Musk is on a downward slope, and from all indications thus far with TruthGPT (viz to further his right-wing ideology), that slide will continue.
 
  • Like
Reactions: bit_user

JTWrenn

Distinguished
Aug 5, 2008
257
171
18,970
Kind of puts a different spin on his signing on to a letter saying they need an ai pause earlier this month. Looks like most things with him, that was just bs to help his business.