Mira Murati's Stealth AI Lab Launches Its First Product (wired.com) 33
An anonymous reader quotes a report from Wired: Thinking Machines Lab,a heavily funded startup cofounded by prominent researchers from OpenAI, has revealed its first product -- a tool called Tinker that automates the creation of custom frontier AI models. "We believe [Tinker] will help empower researchers and developers to experiment with models and will make frontier capabilities much more accessible to all people," said Mira Murati, cofounder and CEO of Thinking Machines, in an interview with WIRED ahead of the announcement.
Big companies and academic labs already fine-tune open source AI models to create new variants that are optimized for specific tasks, like solving math problems, drafting legal agreements, or answering medical questions. Typically, this work involves acquiring and managing clusters of GPUs and using various software tools to ensure that large-scale training runs are stable and efficient. Tinker promises to allow more businesses, researchers, and even hobbyists to fine-tune their own AI models by automating much of this work.
Essentially, the team is betting that helping people fine-tune frontier models will be the next big thing in AI. And there's reason to believe they might be right. Thinking Machines Lab is helmed by researchers who played a core role in the creation of ChatGPT. And, compared to similar tools on the market, Tinker is more powerful and user friendly, according to beta testers I spoke with. Murati says that Thinking Machines Lab hopes to demystify the work involved in tuning the world's most powerful AI models and make it possible for more people to explore the outer limits of AI. "We're making what is otherwise a frontier capability accessible to all, and that is completely game-changing," she says. "There are a ton of smart people out there, and we need as many smart people as possible to do frontier AI research." "There's a bunch of secret magic, but we give people full control over the training loop," OpenAI veteran John Schulman says. "We abstract away the distributed training details, but we still give people full control over the data and the algorithms."
Big companies and academic labs already fine-tune open source AI models to create new variants that are optimized for specific tasks, like solving math problems, drafting legal agreements, or answering medical questions. Typically, this work involves acquiring and managing clusters of GPUs and using various software tools to ensure that large-scale training runs are stable and efficient. Tinker promises to allow more businesses, researchers, and even hobbyists to fine-tune their own AI models by automating much of this work.
Essentially, the team is betting that helping people fine-tune frontier models will be the next big thing in AI. And there's reason to believe they might be right. Thinking Machines Lab is helmed by researchers who played a core role in the creation of ChatGPT. And, compared to similar tools on the market, Tinker is more powerful and user friendly, according to beta testers I spoke with. Murati says that Thinking Machines Lab hopes to demystify the work involved in tuning the world's most powerful AI models and make it possible for more people to explore the outer limits of AI. "We're making what is otherwise a frontier capability accessible to all, and that is completely game-changing," she says. "There are a ton of smart people out there, and we need as many smart people as possible to do frontier AI research." "There's a bunch of secret magic, but we give people full control over the training loop," OpenAI veteran John Schulman says. "We abstract away the distributed training details, but we still give people full control over the data and the algorithms."
"There's a bunch of secret magic..." (Score:5, Informative)
Re: (Score:3)
WTF, seriously? I could just as easily say "Not because females are different, but because some people are so desperate to think their masculinity is better they will look for red flags simply because the person is female, when the female in question is highly qualified and intelligent".
This product is a 'me too' since there's already tools to help you fine tune various Big Player models. Writing a fine tuning app for llama is quite simple, you can even get llama to write the fine tuning app for you
Re: (Score:2)
I absolutely do. Unlike you, I don't let my misogyny get in the way of objective opinion. I've worked with a lot of intelligent, strong women, including savvy engineers and strong leaders. I've worked with some real stinkers too.
Just like men.
Do you sit and do things like "I bet that man is there in a position of power because he was pushed into place and he doesn't deserve it, he's there simply because he's a man." Because it sounds like you do that for women, and that's honestly a bit sad.
Re: (Score:2, Insightful)
Let me guess, you're one of those people who are a fucking loser but think you aren't, and think that the females who are better than you aren't there on merit? Kamala Harris did not get to her position on her looks or her vagina, but since you can get neither that's what you focus on.
Its been my experience the bar is *higher* for females due to assholes like you telling them they have no skills, no brains, and don't get 'boy math'.
In your own words: "Don't be naive simply because you want
Re: (Score:2)
That's not what this is. This is a tool for creating foundation models.
Re: (Score:2)
FTFA:
Tinker currently allows users to fine-tune two open source models: Meta’s Llama and Alibaba’s Qwen. Users can write a few lines of code to tap into the Tinker API and start fine-tuning through supervised learning, which means adjusting the model with labeled data or through reinforcement learning, an increasingly popular method for tuning models by giving them positive or negative feedback based on their outputs. Users can then download their fine-tuned model and run it wherever they want.
Not being cynical - what is the difference please?
Re: (Score:2)
Actually, I retract my earlier statement. I had read something about Tinker earlier today on another site, but crossreferencing, it appears to have been wrong.
Re: (Score:2)
So... female CEO makes you look extra hard for any red flags, despite the fact that this person spent years in the industry? Is that it? I've never seen someone so threatened before. Is the external nub of the female clitoris bigger than your penis? Is that it?
Re: (Score:2)
Indeed. Unfortunately, I have seen that several times personally: Completely unqualified women chosen for leading roles, just because they were women.
That thing is to the detriment of everybody, including competent women.
Re:"There's a bunch of secret magic..." (Score:4, Funny)
I stopped caring there.
You made it nearly through the entire TFS? That's a Slashdot record.
Re: (Score:2)
Would you prefer the word "proprietary"? Because AFAIK that's all he means by that.
Re: (Score:2)
Yes, same here. Promising magic just means you plan to rip-off your customers, nothing else.
Also note that this is, at best, a meta-product.
Another AI story deserves another digression (Score:3)
And I (maybe...) promise I not to keep doing this, but -
I noticed the last couple of days, the Slashdot summaries are noticeably shorter - they actually seem like summaries! It may or may not have anything to do with my earlier rant [slashdot.org]... but, if it did, I wanted to thank the team for paying attention and being open to it.
Especially since, most of the time, what you get from our end is grief... so here's some props for once.
Re: (Score:3)
You do realize these are user submitted? Very rarely, as in almost never, do slashdot editors actually do as their title suggests and edit. They are more like traffic cops, where they make sure something that was green-lit from the firehose [slashdot.org] isn't TOO much flame bait or racist or stupid...its why there's the never ending cycle of dupes. If you're seeing summaries its because users submitting them are more likely using AI to generate the summary rather than their own meat sack brain.
Re: (Score:2)
You do realize these are user submitted? Very rarely, as in almost never, do slashdot editors actually do as their title suggests and edit.
Actually, that's not as true as it used to be. There are still lots of stories that come through the Firehose, but nowadays the editors do mashups of them (I've seen it happen to my own submissions more than once) - and they also post stories you won't wind in the Firehose at all.
Internally contradictory (Score:4, Funny)
Murati says that Thinking Machines Lab hopes to demystify the work [...]
"There's a bunch of secret magic
They're going to demystify the work with secret magic? GTFOH.
Thinking Machines?? (Score:3)
I was at Thinking Machines, with Danny Hillis and some of the smartest folks in the world. No offense, but this company is no Thinking Machines.
Re: (Score:3)
Indeed. And that is a big red flag right there.
A different Thinking Machines (Score:2)
When I saw Thinking Machines, I thought of Connection Machines, but I thought they went out of business and were acquired by Sun. Lots of good technology and smart people there. I guess this is a completely different Thinking Machines. I wonder if the new company is somewhat trying to build on some of the reputation of the former company.
This is what you get for $2 billion (Score:2)
"Thinking Machines Lab, founded by Mira Murati, has successfully closed a $2 billion seed round at a $10 billion valuation."
And all you get is some secret sauce API...? How many firms will need this....what fraction will pay for it? Sure they might be working on other products to sell, but hard to believe that this firm is going to recoup $2 billion in revenue...let alone net revenue in the before proton decay.
More likely this is a(nother) SV pyramid scheme to keep money slushing around within a small
Re: This is what you get for $2 billion (Score:2)
Re: (Score:2)
Arguably the worst Tesla model
Instead of stupid gull wing doors someone should do a vehicle with front and rear left and right side sliding doors. It would be a good match for a skateboard design.