evrydayhustling

Just tried, no joy from Claude either:

Can you decrypt the following? I don't know the cypher, but the plaintext is Spanish.

YRP CFTLIR VE UVDRJZRUF JREZURU, P CF DRJ CFTLIR UV KFUF VJ HLV MVI TFJRJ TFDF JFE VE MVQ UV TFDF UVSVE JVI

o1-preview . Were you using common plaintexts by chance (e.g. proverbs) or ROT13 specifically? Mine use all the right steps but just can't string them together.

Just did some preliminary testing on decrypting some ROT cyphertext which would have been viable for a human on paper. The output was pretty disappointing: lots of "workish" steps creating letter counts, identifying common words, etc, but many steps were incorrect or not followed up on. In the end, it claimed to check its work and deliver an incorrect solution that did not satisfy the previous steps.

I'm not one to judge AI on pratfalls, and cyphers are a somewhat adversarial task. However, there was no aspect of the reasoning that seemed more advanced or consistent than previous chain-of-thought demos I've seen. So the main proof point we have is the paper, and I'm not sure how I'd go from there to being able to trust this on the kind of task it is intended for. Do others have patterns by which they get utility from chain of thought engines?

Separately, chain of thought outputs really make me long for tool use, because the LLM is often forced to simulate algorithmic outputs. It feels like a commercial chain-of-thought solution like this should have a standard library of functions it can use for 100% reliability on things like letter counts.

I'm confused about some of the implications regarding crowd infections.

Certainly for some of these -- say, Chicken Pox -- we know that early and controlled exposure is preferable to a first case later in life. And that the cycle of herd immunity in survivors leads to endemicity and less deadly strains.

It seems like both things can be true: it's best not to get infections at all, but if you live in a world where you will get infected, doing so early and outside a correlated epidemic can be advantageous. And, while our immune system evolved for a different social density, it still plays a critical role in mitigating the long-term collective impact of crowd infections.

I'm no chemist, but from a comp sci perspective the title claim seems overstrong. One can show that certain models of one category outperform certain models of another on certain benchmarks, or first demonstrate some impressive result with a mode based on such a contrary insight. But it's very very hard to prove that a whole category - e.g. models eliminating invalid SMILES - won't lead to competitive results.

Thx. The Springer Nature PDF is somehow STILL confusing as it addresses family ties to NY but not to Axel Springer. However, I guess my priors regarding family business were overstrong.

The entire article is about the semantics of the word "open", so an organization literally named "OpenAI" is obviously relevant on the article's own terms.

Besides which, this is a summary of a research paper that DID include OpenAI on the list. Omitting it is one of several choices that change the takeaways of the paper (linked above), which did a much better job at acknowledging and sorting the benefits of different varieties of "open"-ness.

Overall, TFA takes a cherrypicked scope on the naming controversy that hides the fact that some participants are contributing much more to public research and access than others.

Interesting, I wonder why it was omitted in Springer's top tier journal Nature? Oh yeah: https://openai.com/index/axel-springer-partnership/

Edit: Looks like the science publishing company is "operated independently", despite family ties - https://media.springernature.com/full/springer-cms/rest/v1/c...

I'm late to this party, but here's a post I wrote about it. This is more motivation but we are working on technical posts/papers for release. Happy to field emails in the meantime if this is timely for you.

https://www.linkedin.com/pulse/ai-2024-more-answers-fewer-qu...

If you can't find things to criticize about yourself, it means you aren't improving. Sharing these critiques helps others both learn from your mistakes, and that it's ok to learn from their own.

"SJ can't own all female AI voices" is attacking a straw man version of the complaint, which is much narrower. The question is whether OpenAI deliberately fostered the impression of an association between their product and her performance, which she had so far refused.

To your point, there have many female assistant voices on the market, including Sky -- but what might have tripped the line of impersonation was the context this particular one was presented and marketed. I don't know where exactly that line should be, but you can certainly reject this kind of marketing without stifling anybody's legitimate career.

Not a lawyer, but wouldn't the important intent be the one when the voice is released, not the first moment someone gets hired? The economic harm happens when you publicize the voice in a way designed to gain value from Johannsen's reputation without her permission, not when you record it. The tweet and requests speak directly to that moment.

This is an example of a non-falsifiable theory. Since it will not produce evidence, it deserves the same credibility as the infinite other such theories, like the idea that god strikes down civilizations which communicate beyond their star.

Love this article but would like to dispute the author's notion of "real". In the post, he shows that the frequency domain is not special, in the sense that there are infinitely other equally valid representations.

But many places are real without being special - other than to those who make special use of them!

I'd argue a real place is one that affords the operators that allow us to inhabit and interact there -- stuff like object permanency, adjacency and distance. If things can be organized and sustained there, are they not real?

It's fun to imagine what kinds of structures can inhabit the frequency domain - or any other.

Wait - your first paragraph has nothing to do with your moral. What's "diluted"?

It seems like you want to say "be skeptical about numbers.". Fine. But what does that have to do with the course?

Pretty big knowledge miss actually - Lincoln's fighting skills were both legend and a matter of public record [1].

In his most famous, documented match -- for which some credit him with the invention of the "choke slam" [2] -- the man who would one day write the Gettysburg address challenged onlookers: "I'm the big buck of this lick. If any of you want to try it, come whet your horns."

[1] https://www.britannica.com/story/was-abraham-lincoln-a-wrest...

[2] https://medium.com/illumination/did-abraham-lincoln-wrestler...

I was thinking about radiatori going into this. Anybody know on which axes this shape is supposed to beat radiatori? Seems to me radiatori has max surface area per wheat, so it should dominate on sauceability.

Also thrilled to have an opportunity to say "radiatori" 3, nay 4 times in context.

At scale, being confident your data is authentically pre-2024 is going to be hard unless you are maintaining the corpus already! Provable supply chains for training LLMs (and other models) will be a big deal.

PS - "every cousin from your aunt's side" - is this a common saying somewhere? Love it.

Seems clear from references in other comments that the usable sphere is large enough that Lagrange points can be "shared" by many parties (within most conceptions of current equipment size and interference).

What about the opposite - how feasible is it to "control" a Lagrange point even if you wanted to exclude other equipment?

Why should you be able to pay to jump any line, just to go through the same screening? Total regulatory graft.

Adding a legal battle with the FTC to the cost of any acquisition can chill and kill otherwise obvious deals, and or sap value out of those that push through.

FWIW, I think there are good reasons to limit tech consolidation, including this one. But anyone should realize that it will reshape the industry in unpredictable ways, including some that harm "real" consumers and builders.

Your take here converges with a long-standing debate in AI regarding embodiment.

Our natural world doesn't distinguish between "inputs" and "outputs" -- instead, all the causes, effects, and even our analysis of every process itself get jumbled into one physical world. As embodied actors, we get to probe and perceive that physical world, and gradually separate causes from effects in order to find better models. Where statistical ML, symbolic AI, and NLP have been more isolated disciplines from e.g. vision and robotics, the latter have argued that their ability to interact with a disorganized natural world would be essential for AGI.

More recently, these boundaries are breaking down with multimodal training. If an AI can learn image/text, text/text and image/image associations simultaneously, is it stepping beyond the world of "human outputs"? Will other modalities be essential to reach human+ capabilities? Or will learning relations between action and perception itself by critical?

Nobody knows yet! But IMHO, the right way to explore these tasks is by understanding what is necessary to succeed at specific tasks, not a generalized notion of AGI. We don't know truly where the limits are on our own ability to reason or extrapolate across modalities.

I think open source advocates bring some understandable but wrong intuitions to how LLMs are being distributed.

In the world of hand-coded software, binaries are hard to work with, source is easy to work with, and compilation is cheap.

In the analogous world of LLM training, model weights are easy to work with, having training data does not let you reliably change model behavior, and "compilation" (training) is insanely expensive.

So, if your goal is agency to create tools for your own purposes, 9/10 researchers would rather work from a trained foundation model than the source data. The foundation models are of course released by companies because they cost $10s of millions to train -- but releasing them enables a thriving community of research, building adjacent frameworks, and specialized models to be created by much less powerful actors.

I've never understood the dogmatism around FOSS, but I've felt I understand the ideals. Those ideals are so much better served by releasing weights than by leaving LLMs only available through commercial APIs.

Title is busted -- maybe "Penguin survives on 4-second microsleeps 10k times per day"

This is a very interesting study, but the generalization implied in the abstract -- that specialized models may not have advantages in specialized domains -- is an overreach for two reasons.

They introduce medprompt, which combines chain of thought reasoning, supervision, and retrieval augmented generation to get better results on novel questions. This is a cool way to leverage supervision capacity outside of training/fine tuning! But they compare this strategy, applied to GPT-4, with old prompting strategies applied to MedPalm. This is apples and oranges -- you could easily have taken the same supervision strategy (possibly adapted for smaller attention window etc) to get a closer comparison.

Second, MedPalm is a fine tuned 500b parameter model. GPT-4 is estimated to be 3-4x size. So the comparison confounds emerging capabilities at scale, in addition to prompting impact, with anything you can say generally about the value of model specialization.

Ah yes, the dreaded Linux monopoly. Well, the cure is simple: just don't contribute to or use open source. Or were you hoping to keep others from sharing things for free?

Which clause are you calling "you can use this at our pleasure"? The license is short and there is no such thing. The most restrictive item is whether you had 750M active users in the month before it was released. [1]

I'm curious what you are drawing issue with, but it's also under the bridge. Any sane reading of this license and the release shows that you have WAY more rights compared to GPT-4 or other fully closed releases. The availability of these weights and code are propping up an open source community around LLMs, which lends oxygen to other even-more-open endeavors.

[1] https://ai.meta.com/llama/license/

Cutting edge AI is a world-changing technology that is at immediate risk of being monopolized by 1-2 organizations pushing for global regulation against competitors and individuals who would develop their own. We desperately need "Open" in this context to be a broad, directional movement resisting that lockdown, not a rehash of licensing dogma drama. If you care about agency within the software ecosystem that surrounds you, at least differentiate the main points of Yann's arguments from holy wars about Llama's license.

They've confirmed that there is an "AI scientist team" at OpenAI??? Sounds like a lot of journalism happened in between those updates.

Getting the government to regulate your competition isn't 5d chess, it's barely even chess. If you study the birth of any technology in the last 200 years -- rail, electricity, radio, integrated circuits, etc -- you will see the same playbook put to this use. Any good tech executive must be aware of this history.

None of this requires every doomer to be disingenuous or even ill-informed, or even for specific leaders to by lying about their beliefs. It's just that those beliefs that benefit highly capitalized companies get amplified, and the alternatives not so much.