John R. Lawson

can you imagine sitting in that audience as a three-star flag officer — decades in the service, multiple post-grad degrees and a virtually unequaled level of command expertise — and being forced to watch this dunderheaded dry drunk give an off-brand jordan belfort seminar?

Funny how the guy yelling about professionalism looks like a middle manager at Men’s Wearhouse.

Hopefully someone is around to deport me. The government phone keeps going to voicemail

Fuck you, we aren’t done “with that shit”.

Oooh yeah this is a sweet re-recording.

I’m behind…! Thanks for reminding me. I am half-fuelled by satire these days

Yeah I think that’s my worry too. Most real people not in the ivory towers and bubbles (cough) have thrown out the baby with the bath water at this point.

Oh stop it. Encarta triggers whole-body nostalgic response and I have things to do!

IMO niche markets in tech and research will keep LLM companies afloat. I’m in awe daily of genetic AI and massive productivity increase (without sounding like a capitalist cog 😩). Businesses don’t mind paying $$$ but it is probably OK that edu turns to small fine-tuned low-error models (or books!?!)

There’s utility in narrowing bands of uncertainty at the very worst!

Reminds me of chaos theory’s hard limitation of knowledge, which makes sense given a connection to information theory and exponential growth of entropy, decision trees (waving my hands a lot here. Thanks for the link, looks great.)

Makes sense! Hit me though. If AI used something stochastic like a genetic algorithm to discover something, then I suppose the human creativity is the initial AI as a tool to enact a creative method. But I might argue (badly) humans are doing the same thing via either psychology or brain chemistry.

Getting the most out of an LLM by acting as if it’s human is both fascinating and hilarious, on the other hand.

I pray to the robot gods that all students must have mandatory training in AI ethics, usage, limitations etc. In hindsight, the most regrettable attribute ChatGPT had was appearing like a search engine. When even the devs can’t explain certain behaviours…

May I ask though: why emotional? Is this the same creativity (innovation, divergence) in both science and art?

As an aside, I liked “Everything is a sequence but only some things are a pattern” I read recently, which got me thinking about how art may not convey anything via another culture’s lens.

You know, this all feels like when the staff put the lights on for a few seconds at a nightclub’s closing time. Yeah yeah, we get the hint, we’re drinking up.

But he did, took it down after six minutes but……..

So much of discussion and misunderstand (and anger!) occurs when there’s mix-up with semantics. For instance, if I say “creativity is rearranging pre-existing blocks in new patterns, some of which are improvements”, how many people will want to write me a strongly worded letter?

I appreciate the time you’ve taken to reply with these. You highlight an issue: probability. No LLM tuned to please an average user would risk a paradigm-shifting response until you really shoved it there — like you say, with an expert, breaking axioms etc. I’ll mull over your other comments.

Remember when Brexit to some was meant to stop “them forinurs” coming in, yup those nasty Polish people coming and being European and all, but then ends up increasing immigration from Commonweath countries in global south to fill roles. I laugh when the racists can’t even do racism right.

Honestly, I’m ignorant as a result of discipline hopping and jargon illiteracy, but if I iterate logic with an LLM over new scientific research and it finally makes a breakthrough, is that evidence of creative AI and emergent properties over mere language?

I’ll be all woo-woo and say that LLMs have shown more emergent properties that Shannon would’ve foreseen. I hope we keep prodding and poking the models for these unknown benefits, and for knowledge and not profit. (Can we optimise for both? 😬)

And energy costs 😬 oops, being more depressing now

But… (this is about the third clarification issued!)

I hear that is a Jesus thing to do actually

Oh these are good ones too. Didn’t even see these. Well that’s quite an upsetting one.

That is a disgusting act

I can attest that it is more nuanced than that, but part of the problem is time needed to create a local knowledge base and optimise prompt wording. Might not save some people’s time in some cases? Any generated content should contain an expression of risk: can’t see a way forward otherwise.

I’m building something “similar-ish” and it’s all version systemed and estimates of uncertainty are carried through the “pipeline” of agents. I hope use of smaller models encouraged more explainable, transparent AI.

It is hard to tell if he hates Americans more than foreigners