Steve Gallancy

Father,compute-er,cat fancier,runner,crossworder,reader,writer, sketcher,stretcher,Excel-er,joker,ex-smoker,NewYorker,Florider in no particular order.

[Steve Gallancy @SteveGallancy]

I wonder often about what it’s like not to be curious

[Steve Gallancy @SteveGallancy]

Case study in having someone on the product team who does not think like everyone else on the team

[Steve Gallancy @SteveGallancy]

Considering the frequency with which (non-coding) people using LLMs are asking them to clean up their documents and emails to do just that, it would be interesting to see how these texts might approach the same problem. And substitute teacher Peggy Hill might have written a guide to avoiding the passive tense and closing one’s thesis with what it set out to accomplish!

[Amazing Physics @amazing_physics]

When physics meets art, the result is pure brilliance. 🎨⚛️

[Steve Gallancy @SteveGallancy]

@polynoamial @OpenAI Reasonable question to ask but I rarely see -- can lower tier models solve the same problems, but with more time effort or token spend?

[Steve Gallancy @SteveGallancy]

One of those conversations that will be interesting to look back on in 5 years.

If we're honest, even we zealots know that the eventual transparency about the actual cost of bringing massively parallel computations to bear on question-answering and code generation tasks could bring the whole Jenga tower down.

It will be more and more important to keep the babies (new paradigms for software engineering, semantic search, document decomposition & supercomposition) while the bathwaters are tossed.

[Steve Gallancy @SteveGallancy]

Anthopic, OpenAI Should Not Be Allowed to IPO, Says Ed Zitron https://t.co/K3HPgunPfY via @YouTube

[Steve Gallancy @SteveGallancy]

How about optimism coupled with activism? The winds are still blowing in some bad directions.
Open source initiatives – probably driven and funded from multinational consortiums of companies – could build and maintain models that are good enough for most uses, and they could adhere to principles of transparency with respect to training data, auditable and explainable inference, and user privacy.
Rug-pulling companies like OpenAI and Anthropic won’t bring this: they’ve brought us a long way in short time, but they’ve also brought us a long way from systems that hoovered up all of the data without permission but “that’s Ok, it’ll cure cancer and make every person on earth a millionaire”, and took in billions of investment to counter the prospect that Google achieve a monopoly, to then become creators of a communist-crony-capitalist oligarchy in the West. And China – an actual communist-crony-capitalist oligarchy – has in the most ironic fashion been forced to embrace open-source in order not to be eclipsed. No one (but another crony) would build a business model relying on that.
There’s still a role for very locked-down, private labs like Anthropic and OpenAI to work in niches like cyber and weapons research. Niches don’t merit $2 trillion valuations, of course, so investments and many of the research staff can flow to the consortium-based developers and providers.

[Steve Gallancy @SteveGallancy]

Grok Imagine prompt:

Make the pancakes of this image come alive and flip chaotically from the box onto a nearby plate

[Steve Gallancy @SteveGallancy]

@elonmusk Shotguns need to be refitted to blast out drone-seeking bee-sized minidrones- you’re welcome!

[Steve Gallancy @SteveGallancy]

@satyanadella To paraphrase Steve Ballmer: “Open source! Open source! Open source!”

[Steve Gallancy @SteveGallancy]

@SophiaCai99 "My company's technology will take over all the work for every white-collar, technical and creative job in the world ...and if anything goes wrong, we'll get back to you in, oh, an hour and 15 minutes. ish"

[Steve Gallancy @SteveGallancy]

@AnthropicAI You'd think this might merit some updates over the weekend, particularly when the narrative from the likes of @DavidSacks is that the Gov just wanted you to patch one vulnerability and you refused.

[Steve Gallancy @SteveGallancy]

It’s a bewildering misfortune that Anthropic seems simultaneously to make the best models and harnesses and the worst strategic decisions at the same time.

[Steve Gallancy @SteveGallancy]

Dammit, I have to wait until TACO Tuesday to play with Fable now?!

[Anthropic @AnthropicAI]

The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.

The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.

Access to all other Claude models is not affected.

We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.

Read our full statement: https://t.co/bwn0sximKZ

[Steve Gallancy @SteveGallancy]

My earnings at the Möbius strip club

[Steve Gallancy @SteveGallancy]

The “Workout Exoskeleton” project seems surprisingly under-ambitious. Given that all exercise (cardio, strength, flexibility) aims for salubrious outcomes by provoking biochemical responses that signal adaptive responses, why not try to elicit these directly without the “middleman” of biomechanical movement? Only marginally successful attempts have already been made in terms of electrical stimulation of muscles, and the use of growth hormones and steroids apparently do more harm than good. But here the moonshot would be finding the Goldilocks cocktail/therapy that makes anyone a decathlete – or at least bodily equipped to be one, as the actual skills would have to be acquired through training (until the proverbial Matrix/Kung-Fu skill transfer becomes possible.)

[Steve Gallancy @SteveGallancy]

@Teslaconomics Optimus-Surgeons don’t really need to walk, so if that’s a purpose, why not slash that significant development cost and get these into cheap operating rooms sooner?

[Steve Gallancy @SteveGallancy]

First, I get that this guy is reclusive, but I have to wonder if there’s print with how loudly he’d squawked about paying higher taxes.

Second I’m glad someone else in the comments has covered what a drop in the bucket this is when the annual military payroll is about 200 billion. This covers about 5 hours of payroll 😂

[Steve Gallancy @SteveGallancy]

Why not allow users to optionally first see GPT’s assessment of the best model for the query, allow for accept or reject (in either direction).
Often I want a quick approximate answer to what would seem to be a complex query, and sometimes a deep answer to a seemingly simple problem

[Steve Gallancy @SteveGallancy]

@OpenAI Maybe you can open source GPT-4

[Steve Gallancy @SteveGallancy]

@OpenAI Retiring the old models will make it hard to recall how they would have responded differently to the same prompts.