...which is why we got comically disastrous system-prompt-level attempts to "correct" this once a quarter last year (I haven't kept tabs this year, and most submissions referencing grok "incidents" get flagged off HN quickly, for better or for worse)
I wouldn't trust XAI to refrain from attempting such "alignment" with proper training techniques, in ways that won't result in obvious gaffes.
A manic riff on https://xcancel.com/OpenRouter/status/2065856853989270011 , which advertises https://openrouter.ai/fusion/1 , which is a (slow) multi-model multi-prompt workflow that's specific to the "DRACO" benchmark for "deep research", and doesn't say much about coding and long-horizon agentic work, nor does it imply you can somehow parlay this into duct-taping 50 budget-tier models together for even more gains. Not even sure what "solo" even means in the context of the comparison chart - oneshot? Variant workflow since it doesn't make sense to run on one input?
Mixing outputs of different models one way or another is old news, if it were anywhere near as promising as the author dreams it would have exploded many months ago.
Culinary turmeric is about ~3% curcuminoids by weight (and only 60-70% of that is curcumin specifically). Curcumin also has low oral bioavailability, typically offset by taking a large dose (1000mg) and combining with piperine - even ignoring the piperine, that 1g of curcumin amounts of 66g of turmeric.
The average Indian household does not use 60g of turmeric per person per day. More like 1.5-2g per person per day, or ~30mg of curcumin, and without much to improve absorption.
Curcumin can, in fact, interact with anticoagulants and affect iron absorption at high supplemental doses, which is not a concern at culinary amounts.
There are reasons to be skeptical of the clinical evidence for curcumin supplementation, but "the heterogenous population of India isn't experiencing widespread miracle cures from culinary turmeric" is not one of them.
(And yes, garlic extract is also a thing, also extremely concentrated compared to eating whole garlics or seasoning with garlic powder, and has antiplatelet/anticoagulant activity that one should be aware of before taking such supplements)
"EPA may impede recovery from repeated concussions"
The more pertinent advice would be to avoid getting concussed all the time, but if you're a NFL player, reconsider EPA-heavy supplements (DHA is not implicated).
The framing leads many people to pick blue for its altruistic framing. Enough, in fact, that 50% quorum is honestly not difficult. A lot of red-advocates seem to have a False Consensus Effect going where they're convinced way more people than in reality will interpret this "dilemma" as "do you step in the human grinder in hopes of jamming it", and act accordingly.
A 70% or 90% requirement, or just explicitly framing it as "do you step into the human grinder" would make it vastly easier to aim for 100% red, but we're dealing with the literal words of the "everyone lives button" here.
Clams. Clams and oysters and such. Sessile bivalves are the plants of the animal kingdom, the "genetically engineered brainless cow" of nature. They're also environmentally friendly even when farmed, and more healthy than any animal meat while addressing the same nutritional needs and more. They're almost comically ethical and healthy (and seafood dishes are great imo), they just don't produce bacon and burgers specifically.
When the algorithm is for estimating consumer surplus, the line between coordination and independent cost-optimization disappears.
Why would you try to one-down on price if an “objective statistical AI algorithm” tells you you’d be leaving money on the table without gaining market share to compensate? All it takes is for the market to be sufficiently concentrated at that point.
“Listen to the economists about the economy, not us” sounds reasonable on its own, but the names LeCun lists are all in the lower/modest AI capabilities camp (and there are economists modeling under the assumption of higher capabilities), so it looks like a thinly veiled proxy for more unresolvable bickering over future AI capabilities predictions.
Mildly amusing that "◶NASAFORCE technologists" sounds like a natural enough string in context that it becomes a garden path sentence leading away from that interpretation.
Having looked at https://doublespeed.ai/ out of morbid curiosity, I have to say a simple screenshot would have sent the message more effectively. Well, that and the tagline "a16z funded this".
> The fact you need to work for wealth is a convention of our constraints
The current constraint is "you need to produce to have things".
If one company's AI takes all the jobs, and thus does all the producing-to-have-things, the constraint transforms into "you need that company's permission to have things".
All fiber-consuming gut bacteria, yes - but that's basically synonymous with "good"/beneficial gut bacteria, so it's good advice even if it doesn't give people the massive gainz they might have been hoping for.
I remember when people were discussing the “performance-improving” hack of formulating their prompts as panicked pleas to save their job and household and puppy from imminent doom…by coding X. I wonder if the backfiring is a more recent phenomenon in models that are better at “following the prompt” (including the logical conclusion of its emotional charge), or it was just bad quantification of “performance” all along.
> I strongly suspect that vast majority of the "innovation" in recent years has gone straight to supporting the funding model and institution of the software profession, rather than actual software engineering.
Feels like there’s a counter to the frequent citation of Jevon’s Paradox in there somewhere, in the context of LLM impact on the software dev market. Overestimation of external demand for software, or at least any that can be fulfilled by a human-in-the-loop / one-dev-to-many-users model? The end goal of LLMs feels like, in effect, the Last Framework, and the end of (money in) meta-engineering by devs for devs.
Norway switching from ICEs to EVs objectively reduces global oil consumption+burning by exactly that much.
Norway exporting oil increases oil supply, but doesn't increase consumption. The world's oil consumers are not supply-constrained; the producers are not running at 100% capacity, and they'll happily pick up the slack if Norway just stopped exporting oil for no reason. And there's a large amount of consumption that can't be offset by electrification in the first place (petrochemicals, long distance flight, etc) so there's not even a theoretical future end-state where they require a non-EV-using counterparty to buy their oil to fund their EV usage.
Calling it a "bookkeeping trick" is just verbal sleigh-of-hand.
Because long-term calorie restriction is 100x harder than popping a pill and downing a protein-and-fiber shake, and you can't outrun a burger but you can outlift a calorie deficit, so lumping them all together under "improve diet and exercise somehow" is a nonsensical rhetorical flourish / troll move?
“i can ask it to give a text description of a linear logical math process that has been described in text countless times”
If you think “the tacit knowledge and conscious/subconscious reasoning mix that caused X to write like X” can be meaningfully captured by some 1-page “style guide” like llmtropes, I’m not sure what to tell you. Such a style description would be informed by a soup of reviewers that most certainly cannot write like X even with their stronger and more nuanced observations than what the LLM picked up.
I wouldn't trust XAI to refrain from attempting such "alignment" with proper training techniques, in ways that won't result in obvious gaffes.