cs702

14 ms·

1. 271.▲
   by cs702 1y ago

   Perfect fodder for a horror movie script.
2. 272.▲
   by cs702 1y ago

   See also https://www.reddit.com/r/MachineLearning/comments/1lu1ia0/r_...
3. 273.▲
   by cs702 1y ago

   I would read the blog post by the lead author instead of watching this video: https://alexiglad.github.io/blog/2025/ebt/ Also, see: https://www.reddit.com/r/MachineLearning/comments&
4. 274.▲
   Energy-Based Transformers Are Scalable Learners and Thinkers(alexiglad.github.io)

   4 points by cs702 1y ago|1 comments
5. 275.▲
   by cs702 1y ago

   Yes. I updated my comment to reflect as much. Thank you.
6. 276.▲
   by cs702 1y ago

   The title is biased, blaming Google for mistreating people and implying that Google's AI isn't smart, but the OP is worth reading, because it gives readers a sense of the labor and cost involved in providing AI models with human f
7. 277.▲
   by cs702 1y ago

   Thanks for sharing that post here!
8. 278.▲
   by cs702 1y ago

   "The climate has always been changing." - Lee Zeldin, EPA Administrator Source: https://www.cnn.com/2025/08/03/politics/video/administrator-...
9. 279.▲
   Sorry, Inflation Still Lives(wsj.com)

   5 points by cs702 1y ago|2 comments
10. 280.▲
    by cs702 1y ago

    Direct link: https://www.apple.com/ios/feature-availability/#apple-intell... Live translation with AirPods is not available (quoting) "if you are in the EU and your Apple Account Country or Region is also i
11. 281.▲
    by cs702 1y ago

    Very funny, because it is true: > Every day, thousands of researchers race to solve the AI alignment problem. But they struggle to coordinate on the basics, like whether a misaligned superintelligence will seek to destroy humanity, or ju
12. 282.▲
    by cs702 1y ago

    The Fed's inflation target is a long-term stable 2% annual increase in prices of personal consumption expenditures (PCE): https://www.federalreserve.gov/economy-at-a-glance-inflation... The motivation for 2%/year
13. 283.▲
    by cs702 1y ago

    I hope you're right. To the extent I come across as doubtful and concerned, it is only due to recent firsthand experiences buying food, clothes, and electronics. My perception is that prices are still rising in many product categorie
14. 284.▲
    by cs702 1y ago

    > We’re already in for a world of hurt. I fear you're right, but hope you're not. Yes, this is wishful thinking on my part. I don't want to wish ill on anyone.
15. 285.▲
    by cs702 1y ago

    All I can suggest is to read Buffett's classic article from 1977: https://fortune.com/article/buffett-how-inflation-swindles-t...
16. 286.▲
    by cs702 1y ago

    As he raised the federal fund rate to 20%, Volcker became widely hated indeed. Businesses and consumers suffered because of his actions. Politicians of course tried to take control of the Fed. They also tried to fire him. At one point he
17. 287.▲
    by cs702 1y ago

    You're right. Fixed! I knew that, but my puny little brain somehow didn't catch it as I was writing my comment.
18. 288.▲
    by cs702 1y ago

    If unemployment continues to rise, and the one-time impact on prices of tariffs leads to self-reinforcing high inflation, we'll have stagflation : https://en.wikipedia.org/wiki/Stagflation I sure hope not , beca
19. 289.▲
    by cs702 1y ago

    One country ≠ the world.
20. 290.▲
    by cs702 1y ago

    I disagree: * Vehicle sales represent the bulk (typically, 75-80%) of Tesla's revenues, so I'm not comparing apples to oranges. * A decrease of 12-13% is not a "collapse." That's blatant exaggeration, not "rhet
21. 291.▲
    by cs702 1y ago

    I didn't say anything about Tesla's valuation. Please don't attack a straw-man.
22. 292.▲
    by cs702 1y ago

    Year over year, vehicle deliveries decreased 13%. Quarter over quarter, vehicle deliveries increased 14%. That's most definitely not a "collapse."
23. 293.▲
    by cs702 1y ago

    > The company's sales have collapsed across the world Well, that's news to me. According to Tesla's presentations and quarterly SEC reports, sales have most definitely not "collapsed." In the most recent qu
24. 294.▲
    Americans face biggest increase in health insurance costs in 15 years(ft.com)

    62 points by cs702 1y ago|49 comments
25. 295.▲
    by cs702 1y ago

    Yeah, the masses now decide who is in control of their narrative. Unfortunately, the masses today tend to be poorly educated in history, politics, philosophy, economics, the sciences, and mathematics. Look around. Poorly educated people c
26. 296.▲
    by cs702 1y ago

    Maybe. At the end of the OP, Hinton qualifies his guesses with: > "We don’t know what is going to happen, we have no idea, and people who tell you what is going to happen are just being silly," he adds. "We are at a poi
27. 297.▲
    Interview with Geoffrey Hinton(ft.com)

    41 points by cs702 1y ago|122 comments
28. 298.▲
    by cs702 1y ago

    It's possible. We're talking about pretraining meaningfully larger models past the point at which they plateau, only to see if they can improve beyond that plateau with RL. Call it option (3). No one knows if it would work, and it
29. 299.▲
    by cs702 1y ago

    > Of course we can, ... synthetic data ... That's option (2) in the parent comment: synthetic data.
30. 300.▲
    by cs702 1y ago

    I don't think Sutton's essay is misunderstood, but I agree with the OP's conclusion: We're reaching scaling limits with transformers. The number of parameters in our largest transformers, N, is now in the order of trilli