I'm probably a good candidate to direct a platoon of agents. I'm a generalist and I think in terms of business requirements. But I'm not sure I want to. Coding captures my imagination, and directing agents just plain doesn't.
Or maybe it's more than that: maybe I'm off-put by people who have no need to be in the immediate AI race spending a lot of money to get ahead without asking what near-term problem they are trying to solve. It's depressing and makes the whole field more depressing. My advice for them (if they cared) would be that soon all this will be even more batteries-included, to where any dunce can dial up a production-ready app with a sentence. There's no need to rush; when it happens you'll be better off not having wasted millions trying to be on the bleeding edge.
> they should be more rigorous about carefully defining the knowledge objectives of the class, thoroughly breaking down complex skills into components, and doing lots and lots and lots of practice
> As someone who makes use of AI quite a bit in my own learning, I can say that it’s still relatively weak at having a good model of an individual’s skill gaps and conceptual weaknesses
Why not spec out the curriculum and spec out the approach (regular quizzes, etc.), then use that to guide the AI? Make the skill gap an objective thing.
This was such a good discussion. So many interesting topics hit.
I find his take on open source software interesting; that pushing for free and open software is inevitably serving the interests of those who will collect and profit from it. And that by attempting to make everything free we are making it more difficult for young people to make money. Lots more to say about that I'm sure.
On social media I agree with him that the current economic rules are creating the undesirable effects we see now. My solution for this was to ban the sale and purchase of user data. His perspective was that policing the data was more difficult, and that a better solution would be to police the software; that we should outlaw software that attempts to predict human behavior. There might be unintended consequences there, but it's not a bad idea.
For the GitHub discussion, I don't know how you asked the question, but it would be wise to include in your question what sources you have already consulted, so that they don't also consult the same sources. This is true whether we're talking about AI or Encyclopedia Britannica or microfilm.
Asking questions well is a useful skill that is not at all new.
As someone who finds Japanese corporate culture interesting or even desirable in some ways, it definitely doesn't seem like the most efficient way to run a company. And I'm sure there are plenty of cultural aspects that would not be my cup of tea.
> the J-firm, run by its employees and largely indifferent to the interests of shareholders, exists simply to continue existing
I don't know if all companies should be run like Japanese companies, but there's something very heartwarming about this. Some companies exist for the purpose of employment, and that's okay. In fact it's admirable and makes me want to cheer.
For people, they hear a news report and avoid affected intersections. With autonomous cars they can have people (or AI or both) monitoring affected areas and blacklisting streets and intersections in real time.
> It remains to be seen where we'll converge on capability, incident rate, and acceptance.
I think we're already there with Waymo as the example. We may later choose to diverge from this now-accepted path, but for the moment we have a blueprint, and fixing edge cases with a software update is apparently acceptable, if you just look at all the Waymos operating legally right now.
On a normal day it should suffice to train the model to use its judgment and maybe monitor how other cars are reacting to water covering the road, but when it starts flooding everywhere maybe they should pause the service until it dries out.
Mostly dismissive comments, it seems. Maybe justified. But I think a more interesting conversation is what happens if this or other devices like it become a hit? I wonder if the next generation of users will look at computers with no AI features the way we look at MS-DOS.
Yeah the Pro Pinball series cstarted arriving around the same time as Windows 95. I guess people liked the Windows game because it was just a few clicks away.