Sunday, June 18


Daily News Stuff 18 June 2023

Malicious Incompetence Edition

Top Story

Tech News

  • Another nail in OpenAI's coffin: Falcon LLM has been open-sourced.  (TII)

    It comes in two versions: Falcon-7B which needs 16GB of RAM to run and is suitable for hobbyists, and Falcon-40B which needs 100GB of RAM and is aimed at academic researchers and startups.  Both are available under the Apache open source license, making them free for individual and commercial use.

    Yes, these LLMs are still wildly overhyped, but they're not entirely useless, not when they're free and can run on a laptop - and when they're not intentionally crippled by ultra-woke marketing teams.

    The developer notes make it clear that Falcon was trained on wide set of data from the public internet, making it into a very well-informed digital schizophrenic.

  • AI does not help programmers.  (CACM)

    Or more specifically, it helps bad programmers become mediocre programmers, but it can't go beyond that because Large Language Models do not distinguish right answers from wrong.

    In the example provided, ChatGPT corrects a bug in the sample code, but introduces a new bug.  When this is pointed out, ChatGPT fixes the specific case but not the general case.

    And when this is pointed out, ChatGPT coughs up what appears to be a hastily rewritten version of the author's own work, because the author turns out to be Betrand Meyer, creator of the Eiffel programming language and a leading expert in formal verification of computer programs.

    To be fair, most programmers aren't Betrand Meyer, but my opinion of the current state of AI tools in programming mirrors his.

  • GPT-4 can pass MIT's Electrical Engineering and Computer Science curriculum with a perfect score.  (HuggingFace)

    A remarkable result.

  • No it fucking can't.  (Notion)

    The second paper highlights two problems with the first.

    One, 4% of the problems in the test set cannot be solved with the information provided, or in some cases, at all:
    Below you are given the delays for the different gates you were permitted to use in part D above. Compute the propagation delay of your circuit from D.
    That's the entire question.  There is no part D above, and yet the claim is that GPT-4 answered this question correctly.  There are many questions like this in the test set - this second paper links to a spreadsheet with the full list of questions, good and bad.

    Two, the answers provided by GPT-4 are scored by GPT-4.  If GPT-4 tells GPT-4 that GPT-4 got the question wrong, GPT-4 gets to try again, indefinitely.

    Supposedly the answers were verified manually, but if so, they did a pretty poor job because they missed all the wrong questions.

    Three - not included in the paper, but posted today on Twitter - the original code used to run the tests leaks the answers used for verification by GPT-4 to the GPT-4 instance answering the questions.


  • Why EVs won't crash the electric grid.  (Washington Post / MSN)

    Yes, EVs will require massive upgrades to the electric grid, but the grid has grown at that rate before.

    Well, it hasn't, but it almost did.

    Once.  Fifty years ago.

    For a while.

    We're doomed.

  • Putting a stick shift in an EV because...  (The Verge)

    There is no because.

Disclaimer: There is no because. There is only Zuul.

Posted by: Pixy Misa at 03:32 PM | Comments (5) | Add Comment | Trackbacks (Suck)
Post contains 661 words, total size 6 kb.

1 "Three of the biggest Reddit communities reopened in the funniest way possible"
John Oliver's not exactly funny.  He's a sanctimonious wokescold.

Posted by: Rick C at Monday, June 19 2023 01:40 AM (BMUHC)

2 Yep.

Posted by: Pixy Misa at Monday, June 19 2023 02:16 AM (PiXy!)

3 I've looked over Ed's description, and I tend to think his 'cure' is basically part of the cause.

Talks about 'sustainable', maybe implies socially responsible.

Thing is, when you clutter up the mission of business and executive with all that unmeasureable 'sustainable' crud, you distract from ensuring a profit on every transaction.  If you make a profit on every transaction, and the transactions are voluntary, you can turn that into a metric, and seek it.  It also effectively ensures good, without having to explicitly seek it.

'Sustainable', etc., is unmeasureable where it isn't hooey.  Mix six parts touchy feely nonsense to one part sane business agenda, and you get people managing to metrics like growth, with no idea of what the real fundamentals are. 

Regulatory capture, and employment law basically drive towards randomly hiring work force at overly large companies with no coherent sense of mission or industrial sector. 

If you have a small business, and cannot afford dead weight, at times your hiring will not be matchable to this or that model of what hiring 'should' look like.  This gets you sued.  So, go out of business or join a big group that can hire lawyers and bureaucract or have executives compensated enough to buy influence with the govenrment. 

Posted by: PatBuckman at Monday, June 19 2023 02:24 PM (r9O5h)

4 Fad management - remember that fish market video that kicked off one fad?  Came in to work one morning, and the ceiling had loads of inflatable sea creatures(mostly fish) suspended from the ceiling.  At first you wondered if the place had been converted to a pre-school.  Management thought it was great, workers were annoyed.

Posted by: Frank at Monday, June 19 2023 04:29 PM (rglbH)

5 (Where's Your Ed At)  for some reason that made me think of creepy laboratory type monkeys and I don't know why, losing my damned mind!

Posted by: bob in houston at Wednesday, June 21 2023 12:12 AM (YBLgY)

Hide Comments | Add Comment

Apple pies are delicious. But never mind apple pies. What colour is a green orange?

55kb generated in CPU 0.0151, elapsed 0.5277 seconds.
58 queries taking 0.5187 seconds, 340 records returned.
Powered by Minx 1.1.6c-pink.