an especially nasty surprise when you've used CC a bunch for coding and then claude.ai won't answer your general deep research questions. your context-switching is not Claude's context-switching! 😂 😭
Time to go Through the Looking-Glass…
“The name of the song is called 'Haddocks' Eyes' [..] The name really is 'The Aged Aged Man' [..] The song is called 'Ways And Means' [..] The song really is 'A-sitting On A Gate'”
I don’t think that’s correct
I attended a “fireside chat” with Emil Michael a ~decade ago while he was in charge of Uber’s growth.
His views then on the government’s role in influencing private enterprise policy was… quite different.
what a project to be involved with! 😊
The Dead Flag Blues from Godspeed You! Black Emperor’s F#A#∞
"We're trapped in the belly of this horrible machine. And the machine is bleeding to death"
The irony of this conversation happening with USG, given Amodei’s very public writing about the risk of AI-powered CCP authoritarianism!
We installed a simple device that conveys sound from underwater to above ground, and vice versa. Now the wolves gather at the shore to howl to the whales, and the whales sing back to the wolves
I always regret not taking that train (or the one up to BC) when I was in Seattle.
Sadly I don’t think it hugs the Pacific coast? Could have been one of the most beautiful train rides on the planet!
(Eat your heart out, GWR past Teignmouth… 😀)
if you don’t mind me asking, what do you think of the BBC’s Verify bits?
"Old Man Yells At Cloud" image from The Simpsons
whoever trained the RLHF reward model to like emojis has a lot to answer for!
Can't shake the feeling there's an analogue here to the psychology of goal setting for humans (e.g. www.tandfonline.com/doi/full/10....)
Process goals > performance goals > outcome goals
(and a hard agree on the code smell of default assistant personas. Everyone loves emojis, right? ugh! 🙂)
In fairness, I think Isomorphic Labs (read: Alphabet’s commercialisation vehicle for AlphaFold, etc.) have already put a similar business model in place with pharmaceutical companies.
@den.dev - warm congratulations on the new gig!
Hope you're well... catch up call soon?
how on earth do you find the time? 😆
John, thanks for the enjoyable trip down memory lane :)
Fortunate to still have a memento of that time sitting behind me!
Looking forward to Part 2...
Porkbun
simply acknowledging that Obsidian Sync /is/ a great, albeit, paid solution! But that I’m still sheepishly asking for a free solution…
A core “enrichment” plugin that augments existing data in a vault: identifies entities (people / organisations) and gives the option to creates notes with background information / links to external sources.
Native connectors to standard internet calendar and contact services.
A free, reliable solution for macOS / iOS / Windows vault synchronisation (ack Obsidian Sync 😞)
A native way to use voice input shortcuts to append thoughts to my daily notes
Greater flexibility to use paragraph / heading level metadata as an input to bases / dataviews (cf. file-level frontmatter)
Dafydd Jenkins having a great 2nd half for Chiefs
plus ça change, plus c'est la même chose! 😊
Congratulations to the team on this announcement, Frank!
[1/9] Excited to share our new paper "A Pragmatic View of AI Personhood" published today. We feel this topic is timely, and rapidly growing in importance as AI becomes agentic, as AI agents integrate further into the economy, and as more and more users encounter AI.
I suppose some folks might see this as being analogous to it learning some form of world model.
I agree with your point about people misinterpreting the nature of the problem.
But it’s interesting to consider whether the training algorithm allows, in theory, the model to learn enough of a token’s properties to answer any question about it /as if/ the model could read each character.
It’s feasible an LLM could learn how many of a specific character appears in a token.
When you ask “what is the capital of France” and ChatGPT says Paris [72872], it learned that. It didn’t look up the answer on a map.
So it could learn there are 2 b’s in blueberry. Even if it doesn’t count them.
I love the photos I get from my X100F but I just ended up wanting something I could put in a pocket. So I now carry round a Ricoh GRIII which is fab (but also annoyingly expensive and has a crap battery :/)
I’m target market for the X Half but even for me it’s a little… too quirky 🙃
I know that. You know that. Does the President know that?
