HarHarVeryFunny
HarHarVeryFunny on Hacker News
Is this looping tough to fix by changing the training?
Dario will send you a postcard from Fiji if they achieve RSI, AGI, any mass extinction events, or any other milestones.
AI is now capable of, and being applied to, find so many bugs recently that infrastructure providers need to up their game and respond quickly.
Hopefully this is a passing phase as bugs are quickly patched, and we'll get past it without too many incidents.
In contrast, back when RedHat went public they reached out to everyone on the open source mailing lists, and gave everyone a few shares as thanks for the work they were benefiting from. I got some shares even though all I had done was post some free video format converters.
One of the ways they made this work was by assigning temporary powers to elected/appointed individuals appropriate and sufficient to get their job done, whether that was provincial governor, tribunus plebis, consul or whatever. A notable title was that of "dictator" (where the modern word derives from) which gave sweeping temporary powers to a general sufficient to execute a war.
Julius Caesar famously abused this system, effectively becoming the first emperor (despite no such position existing) by assuming the title of "dictator perpetuo" (perpetual dictator), although that did not end well for him. His successor, Augustus (whose adopted name later became the name of the office of emperor) initiated a more acceptable version of this by simultaneously holding multiple "elected" positions thereby having the power of them all.
These are averages, but still!
Normally, as well as food sources, your body naturally produces Vitamin D as a reaction to sunlight (being outdoors), and not having sufficient time outdoors, especially in a pleasant/nature environment, may be contributing to depression.
If you ask an LLM to "win like Magnus" it can't because, as above, it is a feed-forward auto-regressive predictor. Do you realize how much different this is to how something like Stockfish (that can play at, and far beyond, Magnus' level) works?
A big LLM is still an LLM - the architecture and operation doesn't change. It just has more capacity to learn, both to learn deep predictive patterns and to memorize factual knowledge.
Does it not occur to you that your beliefs of what LLMs are capable of (generating amazing prose) is totally at odds with what we see them doing? Where is all the amazing prose?
Your rewritten codebase will presumably look much cleaner and more maintainable that what it's aiming to replace, but it's easy to miss the purpose/necessity of all of the old "cruft" which need not be just poor coding but rather later enhancements not anticipated by the original design whose implementation was therefore crufty.
Does anyone here work at Cisco, and have any insight as to why a rewrite is considered necessary and might be expected to succeed?
I think both music and humor are both exploiting the same evolutionary essentials of human nature - to find comfort/safety in predictability, and excitement/danger in unpredictability. Laughter would seem to be a social thing - perhaps it helps develop a theory of mind in signalling that you get the joke - that the punchline is indeed unexpected to you as well ? And/or perhaps it helps draw others attention to an unexpected situation with learning potential.
The "quality/goodness" (likability) of output is only selected for to extent that it also reflects RLHF preferences originating from low paid contractors who are scoring based on "looks good to me", not based on Pulitzer prize prose potential.
To explore a false hypothetical, suppose an LLM was only trained on the output of a single popular author, then you'd certainly expect it to reproduce that author's style, but it would not be "upredictably interesting" but rather predictably quirky, reproducing the same "surprises" over and over so that they no longer are surprising and therefore interesting.
To continue the hypothetical, if you trained an LLM exclusively on the output of multiple good authors, then rather than get the best of each you'd get a statistical blend (this is how training works) that emphasized what they had in common rather than the individual quirks that made them interesting.
There is a fairly common misperception that "AI" (LLMs) is always trying to do it's best, where "best" has been anthropomorphized to have the human sense rather than the LLM sense of "the most statistically accurate continuation prediction" it can make. For example, if you give an LLM the moves of a partial chess game and ask it to predict next moves for one of the players, it will NOT play to win, but will rather play to predict how the game would have continued, which means continuing at the level (ELO rating) and style of the player who had made those moves.
Of course you can ask an LLM to play to win, just as you can ask it to write in a style that is, say, "unpredictable and interesting", but it can't escape itself - you will necessarily get predictable unpredictability.
Which requires a bunch of things an LLM lacks, such as theory of mind, knowing who it is talking to, and willingness/ability to engage in short turn conversation to assess and steer the conversation according to the feedback it is getting from the person it is talking to.
It's not just predicting content but also the form/shape of the response, highly shaped by RLHF preferences, which seems to encourage longer form responses which need padding with filler to generate the "preferred" amount of verbiage.
It seems most of these models have been trained to interpret questions as wanting a long-form authoritative/comprehensive-sounding response (even when midwit filler is all it has to offer), rather than initiating a short-turn conversation which would better allow you to steer the conversation. Apparently Muse is much better in this regard, but so far I've resisted signing up for any Meta/FaceBook service, and not inclined to break this winning streak!
Hard to trust that a for-profit benchmarking company isn't biased due to profit-maxxing.
Given that they are now saying they have 25% too many employees, perhaps they should be banned from hiring H1B's altogether until they're rehired however many they end up letting go ...
Except Trump would probably be praising Lukashenko, and admiring his ass, if he was aware of his existence.
I guess if/when we discover a theory of everything then we could claim "this is real/reality", but what would that even mean other than modelling reality/outcomes with 100% fidelity?
Perhaps we've got it backwards and quantum entanglement is what defines "locality" rather than mysteriously existing in the face of it.
Many people of course, Feynman obviously included, understand the theory in the sense of knowing how to correctly apply it to make predictions.
Apparently Ukraine does use already use Eutelsat in addition to Starlink, but it's utility is limited due to number of satellites and bandwidth, as well as the current weight of terminals (presumably an issue for weapons-borne receivers).
https://www.wsj.com/world/trump-called-musk-and-asked-him-to...
Of course from Europe/Ukraine's POV, it doesn't make much difference whether it was Musk or Trump/US military that did it. What matters is that a former ally and NATO partner can not be trusted, and eliminating American technological and defense dependence is therefore critical for regional security.
I don't think you can say the "humans are tribal, emotional and irrational" consensus is primarily coming from corporate sources like linkedin - most people would agree with these, both from first hand experience (team sports, racism, wars, etc) as well as having read such analysis.
I think best mental model of what an LLM does, statistically generating continuations word by word, would be a football stadium full of people, each with a keyboard, and a giant jumbotron screen they can all see... A question (cf ChatGPT user input) gets displayed on the screen, then everyone types on their keyboard what the first/next word of the response should be, and majority wins with that word added to the screen, then they all vote on the next word, etc ...
This is really a very close analogue to how an LLM works, and it would be weird to describe the output appearing on the screen as the "take" of the football stadium itself - it's the group statistics of all the people (cf training data).
I doubt there are very many companies that an LLM could start/run outside of the obvious few like custom T-shirts.