wrangle enjoyer ⏹️ ⏸️

5K posts

wrangle enjoyer ⏹️ ⏸️

wrangle enjoyer ⏹️ ⏸️

@Idgitbhh

Entrou em Aralık 2024
194 Seguindo61 Seguidores
Tweet fixado
wrangle enjoyer ⏹️ ⏸️
@CorleyAaliyah @princessakano A lot for ppl in labs literally believe super intelligence is possible, they can create a machine god, and they are actively trying to make AI better than them at AI research If an AI is better than humans at improving itself, you should be very worried.
English
2
0
5
984
Rigel
Rigel@RigelOrionBeta·
@Idgitbhh @TrueSlazac @electionsjoe The point is that we will never know. The best that prediction markets tell us is what the public thinks the probabilities are. Not what the probabilities are. There are incredibly few things anyone can predict with any degree of certainty.
English
1
0
0
4
Joe
Joe@electionsjoe·
Gotta love prediction markets lmao
Joe tweet media
English
34
66
2.8K
150.2K
i/o
i/o@avidseries·
Musk must have realized that if Twitter deboosted certain Third World regions, a lot of the red meat pro-MAGA and far-right propaganda that he enjoys (like that from Ian Miles Cheong and Inevitable West) would get less circulation.
i/o tweet media
English
46
73
715
20.4K
Martin van Staden
Martin van Staden@Martin_ASFL·
@CBHawkeyes @avidseries People work in advocacy to get paid. I got paid to spend hours of my life constructing the Index of Race Law, which is why Elon Musk can say "there's more race laws today than under Apartheid". Me - or the accounts spreading awareness - getting paid, is not a bad thing.
English
2
0
6
118
wrangle enjoyer ⏹️ ⏸️
@RigelOrionBeta @TrueSlazac @electionsjoe purely selfishly, if you are thinking of investing in the US, you now have to take into account the risk of your business + the risk of political instability don't disagree lots of it is basically pure gambling (especially lots of sports betting) but nevertheless
English
1
0
1
5
wrangle enjoyer ⏹️ ⏸️
@RigelOrionBeta @TrueSlazac @electionsjoe I'm saying on average. Like, if you took 100 random bets that are trading at 20%, and you *also* found people willing to bet against you at 50% (1:1 odds) for the same thing, you'd gain money from betting on the Polymarket side (even if you lose in this particular case)
English
1
0
0
12
wrangle enjoyer ⏹️ ⏸️
@Fatwang69420 @tenobrus @cxgonzalez I really do hope you're right, but I also can see with my own eyes the complex (to me) project I couldn't vibe-code 4 months ago, that I am now able to, (so I can stop paying for a SaaS) and that makes me much less likely to believe what you said about Jensen
English
0
0
0
21
J$
J$@Fatwang69420·
Benchmark metrics are irrelevant once you take the outside in perspective. LLMs economically are clunky and inefficient engines chewing tokens to give garbage out, just to consume as much tokens as possible even if you give explicit commands for reduced output. Token sellers are incentivized to not solve the problem quickly 🤣 literally Jensen the other day was talking about a $150k or $250k figure and that’s not from how developers should be using AI but rather how much money they want to make. That’s why the units are in $$$ and not something else. We actually do not have the processing power for 500k engineers to spend 50% of their paid income in tokens but the guy who said they do sells the machines. Jensen is fundamentally looking like he’s subtly realizing this whole AI thing is not going to pan out the way they’ve indicated to investors over the past 2-3 years. It’s a fallacy that you’re going to have Super Devs who do double the amount of work for less than double the amount of money. Notice a shift in rhetoric of “you won’t need devs anymore” as we are up against the limitations of LLM tech.
English
1
0
0
12
christian
christian@cxgonzalez·
LLMs will lead to productivity gains and new forms of high skilled labor, not mass unemployment
christian tweet media
English
108
50
510
22.7K
i2dt
i2dt@SirWirtz·
@TrueSlazac @electionsjoe Unironically the fact people have such a hard time grasping this concept of probability is why it’s so easy to make money on these sites
English
1
0
3
275
wrangle enjoyer ⏹️ ⏸️
@armadillofan23 @TrueSlazac @electionsjoe The results not matching the odds doesn’t mean ppl are acting on inaccurate information (or accurate if they match) The only thing it “proves” is that people are acting on incomplete information, which is true of any predictions ever made (or even any decisions for that matter)
English
0
0
0
5
wrangle enjoyer ⏹️ ⏸️
@armadillofan23 @TrueSlazac @electionsjoe You can have a super biased pollster that decides the US presidential election based on a coinflip and gets it right 10 times in a row You can also have a survey of half of all americans that predicts D, and get ultra unlucky and the other half votes for the R candidate
English
1
0
0
8
wrangle enjoyer ⏹️ ⏸️
@RigelOrionBeta @TrueSlazac @electionsjoe You *are* gaining information. Polymarket is very well-calibrated: events predicted to happen 20% generally the time happen 1 in 5 times. This is predictive. If somebody else makes a 1:1 bet against you, you get money by choosing the side polymarket is on
English
1
0
0
13
Rigel
Rigel@RigelOrionBeta·
@TrueSlazac @electionsjoe This is just tautological. We are not gaining any information. A prediction market is not predictive. It's entirely descriptive. So they're not probabilities.
English
3
0
12
273
wrangle enjoyer ⏹️ ⏸️
@Fatwang69420 @tenobrus @cxgonzalez > LLMs are going to plateau sooner rather than later. why do you think this? LLMs aren't being trained on text alone anymore, they are doing RLVF instead of RLHF Like no benchmarks tests for LLMs seem to be plateaud so far
English
1
0
0
25
J$
J$@Fatwang69420·
@tenobrus @cxgonzalez LLMs are going to plateau sooner rather than later. It’s so overhyped computer science majors can’t explain their product to non-cs majors; complicating every simple task for people to “train a model” and all of these devs live in an echo chamber, zero multidisciplinary care
English
1
0
1
48
wrangle enjoyer ⏹️ ⏸️
@SchopenhauerOn @tenobrus @cxgonzalez this seems wrong bc the stochastic parrots ppl are clearly just wrong (models are being trained to achieve goals now) for RSI: 1) No principled reason AI research is fundamentally different from chess or Go 2) companies have the explicit goals of developing an AI researcher
English
0
0
0
17
a good name for a rock band
a good name for a rock band@alcibiades5410·
@nikicaga You're comparing the entire agricultural process to only the final step of the LLM process, do you really not realize how silly that is?
English
5
0
23
3.1K
wrangle enjoyer ⏹️ ⏸️
@TrueSlazac @stpmtk @ordinarytings that's if you agree with them the chance is that low 😭😭 but even if u think superintelligence is possible and controllable, it matters who's controlling it Like why would Sam Altman ever let us be uploaded rather than be his gardener slaves in his California-sized palace yk
English
0
0
1
21
wrangle enjoyer ⏹️ ⏸️
@stpmtk @TrueSlazac @ordinarytings like you guys, the AI 'ultra-skeptics' (given AI skeptics are way less pessimists than you), should look at the METR graph, which has been increasing exponentially. Ppl keep saying it's an S-curve, but we don't seem to be reaching the flattening part x.com/Idgitbhh/statu…
wrangle enjoyer ⏹️ ⏸️@Idgitbhh

@xereeto @lynettepillled @ChadNotChud mostly agreed, but I'd like to add that AI *is* hitting a wall

English
0
0
0
28
štěpán
štěpán@stpmtk·
@TrueSlazac @ordinarytings so what are the arguments the bubble isn't going to pop? also you should really look into opinions of major AI scientists like LeCunn before formulating opinions on plausibility of superintelligence in the next few years.
English
5
0
3
818