jonmc12

Also surprised; building something people want and proving it is the unlock. HN first principle since the beginning.

I don't think trans-disciplinary inquiry is arrogance - the intellectual fields are somewhat arbitrary relative to how human expertise relates to real world problems. But, effective trans-disciplinary inquiry requires awareness of philosophical commitments, and familiarity with existing literature/theory.

The bigger challenge might be that people with ML expertise need to solve problems of human-AI interaction and alignment because the training for the former is uni-disciplanary while the latter is trans-disciplinary.

Maybe an investment in "A-players" for streaming stifled cultural diversity and kept engineers from being able to innovate on novel media formats where they are losing engagement of the younger demographic to TikTok and other social media video formats.

The same corporate strategy and culture that hired "A-player" engineers for streaming is hiring "A-player" studios for content.

Defining A-players as such means you've set the rules of the game instead of building a culture of adaptive success criteria to meet customer opportunities. The label itself is a function of organizational ossification. This is the likely legacy of our tech giants; innovative in only one direction and not able to change fast enough to avoid becoming a brittle, mediocre institution over time.

As consumers, we can all feel this ossified mediocrity every day.

Agreed. The constraints of software engineering are mostly idiomatic. I used to use my "Scribe" mind to crawl through library dependencies for days to solve some artificial sub-problem.

No software engineer is good enough to time-efficiently write the whole stack from machine code up - it will always be an arbitrary and idiomatic set of problems and this is what LLMs are so good at parsing.

Using "Scribe" cycles to define the right problem and carefully review code outputs seems like the way.

The Enlightenment produced free speech and reasoning. Nietzsche said, "god is dead," but a lot of people said it before and after - because reasoning could not fill in the gap of a shared reality. Harari's Sapiens gives a good history; Hoffman's claim that "natural selection does not favor veridical perception" says you're pretty confused about what is actually going on wrt "truth"; Seth's "Being You" might help to understand what conscious beings are actually trying to do in relation "truth" and survival.

Some of the lower-hanging fruit in chemistry data was addressed by earlier versions of deep learning, like AlphaFold. It seems the nature of the domain is such that the language is less ambiguous than most natural language. Does anyone have a perspective on the apparent advantages of mapping chemistry interactions to latent space models for LLM training?

I think no one has posted the main thesis of this discussion. It's not just about "asking them", it's about the NYU psychologist's research that has systematically found no evidence for all counterfactuals other than asking a person about their internal state.

Knowing how to ask someone what they are thinking/feeling is a key skillset of anyone building a product for someone else. Its nuanced enough that books like "The Mom Test" break this down for entrepreneurs to implement tactically. On the other hand, West's research also suggests that one can comfortably underweight their own instincts laden with ego-centric and culture-centric biases. Further, you can also comfortably underweight the observations of your colleagues who might assert empathic abilities.

Perhaps the most interesting segment of this podcast was the story of how the author and her tenured colleague were able to dismiss their own intuitions about a acrimonious rivalry with one another and evaluate their relationship scientifically through hundreds of questions from their own research. They went from disliking one another to getting married.

As another example, Amazon teams communicate product launch requirements via a future press releases including a FAQ (per description in the book "Working Backwards"). Its a communication intended for the masses with a built-in disambiguation addendum.

Our natural languages uses incremental inquiry to disambiguate context as opposed to using strong protocol. In "Working Backwards", it's the communicator's job to solicit questions from co-workers via pain-staking detailed reviews in meetings ("Bezos scrutinizes every single sentence"). I think of it like constructing a representative survey of ambiguity, and then putting answers in the FAQ that help increase clarity. The more detailed and representative your survey, the more helpful your questions/answers will be to communicate nuance.

With regard to disambiguating through protocol, Organizations evolve jargon to increment protocol, which probably increases semantic alignment somewhat as group size scales. If you read about the history of language, the Rebus principle created protocols of formal alphabets; protocols like grammar gave us formal writing rules. Protocols like TCPIP let our computers talk. Protocol creates more rigid commitments for communication, but also increases potential semantic alignment. As a thought experiment, if we learned to dynamically and deliberately develop jargons en masse, it might create the channels to disambiguate context and communicate nuance at scale.