Plus you can alert the crew if something's off before take off (eg. loose/missing screws/panels). A forner colleague of mine did that, and got to see the entire maintenance crew come out for an impromptu RCA and incident response.
In math, expect it to ne a mess, mostly because nobody is worry about industrial espionage[1]. Some researchers may get an AI version bought by the university, or school or department, others pay from a grant or their own pockets, other use the free version.
I still remember 2020. We (I, my wife and a few close friends) got in total like 5 Moodles for math inside the same university, and I'm not counting the Moodles for physics or other subjets, and probably there were a few more for math that I didn't know. Each one has a slightly different configuration, so it was a mess to teleport info.
Same with email, we have a different email server at the university/scholl/department level. (Who knows how Gemmini/Copilot/Whatever is configured in each one.)
[1] I can't find the source now, but for the first https://en.wikipedia.org/wiki/High-temperature_superconducti... the forced the journal the permision to select their own referees and send the draft with the wrong atomic element and changed it to the correct one at the last printer proof.
> Personally I’d like to see them actually start benefiting humanity by doing all the things Sam has claimed they will like curing disease, cancer, global warming, etc.
It makes more sense to leave curing disease & cancer to the experts, with tools (like AI) being developed by AI experts.
Call me crazy, but I want separate organizations and experts for medical vs finance vs space vs climate vs AI research.
What the op was pointing out is that guys like Altman and Dario are repeatedly saying they’re going to cure xyz diseases and solve xyz huge global problems. Maybe their companies will eventually do these things, but haven’t yet.
I don’t have an opinion either way, I think it’s too soon to tell if llms will be able to cure cancer or whatever. But at the very least it will be a good tool to help researchers do their jobs.
The thing is... AI is not going to solve any problems. People needs to solve their problems. AI can give us clever solutions, but its up to us to do it!
> Maybe their companies will eventually do these things, but haven’t yet.
I think they are working with customers to improve the LLMs and tools for these use-cases. They almost certainly also hire experts to help filter out nonsense, pseudo-science and help curate trusted knowledge bases for training, but it will almost certainly be the customers who deliver the major results, and the AI companies will claim some of the credit. That said, patents for important medicine might help with the bottom line, so I could imagine partnerships and JVs.
> at the very least it will be a good tool to help researchers do their jobs.
> Perhaps there is also "firewalling" on the router and the phone can't even have incoming connections. Not sure.
Correct. Every home wifi router worth its salt will firewall incoming connections by default, whether v4 or v6. It's then possible (unless it's some shitty ISP-provided locked down device) to add specific allow rules, or allow all for a particular client.
egress is typically wide open, although sometimes they lock down particular protocols by default (eg. smtp, bittorrent)
I believe the 1M context window can be configured in Codex (config file edit). The price per token increases when >256k though, but it can be useful when compaction at-that-moment would be detrimental.
Very true. If things are too reliable, systems can come to depend on them always being so so reliable... So it can actually pay off to inject transient issues deliberately.
You know, one of the things I miss from when I wrote most code manually (now it's a minority of code) is that I would have a high rate of random error.
This was extremely tedious, and annoying, of course, but things would break in all sorts of educational ways.
And once in a while, the computer would do something truly magical, something I didn't even know computers can do.
LLMs still make mistakes, of course, but they make different kinds of errors than I do.
I've been trying to replicate the effect with software that reproduces my own type of errors, randomly omitting tokens from the program. (A strange type of fuzzing the program space.)
Some don't compile, some crash at runtime, most don't do anything interesting, but I know for a fact the process is worthwhile.
What would be really interesting would be to create a cheap verification process for this at scale, i.e. a "wall of TVs" each containing the output of a program, so that the "worthwhileness" of each one can be evaluated at a glance.
This is very subjective. You are willing to outsource one core competency (coding), but not another (documentation and other written comms) ?
I think the bigger personal challenge is to not outsource thinking, understanding and judgement as we leverage new tools, and to continue our personal learning journeys. After 20+ years as a software engineer, you can probably coast on past experience for a long time... but if you aren't learning and exercising (one way or another), your skills will stagnate. That would be sad.
reply