@GenCogAI and teaching AI for PM @UCLA prev: @revenue_io | @lynxdxlabs | @dealpath | @belkin | @gyroscope_app | @eff | @ucberkeley | @umsi | @umichlaw
San Francisco, CA
Joined January 2009
- Tweets2.8K
- Following1.4K
- Followers395
- Likes2.9K
I'm fascinated by the use of microblogs to register a bet (by simply tweeting "I predict that..." or "Registering a bet that..").. Despite the existence of cryptocurrency, prediction markets, biometric auth, etc. None of that is needed: you can just tweet things!
All that fuss about web3 decentralization, clever ways to prove if someone is really them, market-making, etc. But in the end, the simple gain or loss of status points based on someone's predictive power is as simple as printing something onto the timeline and waiting a bit...
The least amount of interaction with AI will at some point become a major focus of human status competition
Spent last week at AI Engineer in New York, where the focus was finance. Whatever happens with AI at the big-picture level, the people I met working to deploy what we have today do care about getting it right. A lot of discussions around accurately grounding firm-specific knowledge, thorough measurement and evaluation (the "eat your vegetables" of AI), new techniques with harness engineering and context engineering, and open-source alternatives emerging against 'Big AI'. Have been sharing these insights with my @GenCogAI clients and @uclaextension students alike. Thanks to @vibhusapra and the @latentspacepod community for having me there.
Real ones know
I will say what people in Silicon Valley are afraid to say right now.
It is objectively irresponsible to use a product like Muse for handling anything of consequence or importance. The product has no credibility (zero) when it comes to maintaining one’s data privacy. Instead, the product was obviously been designed to suck as much user data as possible. And all of this assumes the product even works as marketed which is showing to be a bad assumption to hold.
the feeling of being bound by the limits of human understanding, in the face of strong compressive forces
I know it's been 6 months but I'm still thinking about how a once-beloved shoe brand became an AI infrastructure company, practically overnight. Yes, I am still struggling with the Allbirds shoes-to-GPUs pivot. Just think about it: humans are a bipedal animal with two feet, and most of us aim to protect those feet. Hence, shoes are some of the most common material objects that exist in human life. And sure, by the early 2020s, Allbirds had fallen out of style. But for one of these most common human objects to be so subsumed by AI, so quickly, portends something much deeper and darker. Amidst talk of ruinous scenarios, this example unfolding right before our very eyes, might be the real canary in the coal mine. It’s hard not to marvel at how quickly the machine-process of techno-capital can reconfigure a shoe maker into a GPU farm. For a long time (too long) I wore those shoes, on my human feet. And now, conceptually, those shoes transmogrified into silicon chips. Those wool shoes... became silicon chips... to produce intellectual work through matrix multiplications. Is AI awesome? Yes, in the 'I am in awe of how quickly it is destroying common artifacts from my human life' sort of a way. And sure, not everything stays the same.. evolution is an ongoing process bigger than us. But this transmutation is of a different scale and scope than I’ve witnessed. It's not like “Famous shoe company pivots to become tire company”, or “Well-known apparel brand pivots to become photography studio”, or something more coherent like that. This is more like “Something from since before the dawn of human civilization (shoes are between 30,000 and 40,000 years old, meaning they predate civilization itself by thousands of years), becomes a machine-brain farm, in less than 1 year.” I am still trying to fathom what has happened here, and what will continue to happen.
"what are you worried about? the future is so abundant with tokens for everyone"
My @OpenAI ChatGPT account was deactivated for "violating usage policies." I appealed. They upheld the decision with no explanation and told me not to appeal again.
I'm a CS PhD student who works on LLM agents. This looks like a false flag. Can anyone at OpenAI take a look? Happy to DM details.
First, "imagine". Then, "We've been consulting with an independent advisory group”. Finally, a repo drops with the badge "Humans required: No"
Fleet Week? Ah yes, when they valiantly scrambled fighter jets to finally get rid of those annoying planes visually polluting the skies of SF with startup ads during Tech Week
Yes. I increasingly find myself telling my students "No one is going to save you. No one is going to organize on your behalf." Discussed at length by @luke_drago_ here: intelligence-curse.ai/
Insightful talk by @MVXMXM on measuring how much "effort" AI agents expend during tasks. Watt had to invent horsepower in ~1780s to sell his engine against the horse gin, and now we're grappling with this question again for digital work (slide from the talk; link in reply)
Excited to have my Mousepower talk live on @aiDotEngineer's YT channel (link below), but I have retrospective thoughts on this topic & here is what I'd change if I gave the talk again today:
1. Thesis wasn't stated simply enough: value of agents is limited by our ability to measure them, otherwise we can't justify the ROI of their token cost
2. Need an explicit connection to the discourse on verifiable domains, which we need not be limited to today's set as understanding customer mental models unlocks new ways to verify value (e.g. horsepower)
3. After reading a nice post by @jon_stokes on verifiable domains, I'm specifically interested in the n-order F/X of their "optimization pressure"
4. Also rambled too much, perhaps a function of the topic being too underbaked... will see if GPT voice can coach me for next time
Follow up incoming (someday).
