spacemanspiff01

There is all of this talk about an ai winter, and somehow I just kind of doubt it. Unlike in past decades there is a ridiculous amount of money, energy, and time going into the space, and unlike 40 years ago (or even 10) there are things that can actually make money now, and are not just academic research.

(See stable diffusion/llama/ chatgpt)

There will be businesses that actually make money on these technologies, and they will be research heavy (even a 5% improvement is a big deal) as things are still getting figured out.

I could see speed dropping back towards 2017 like rates, but I kind of doubt we will ever see an true ai winter like the 90's early 00's

The field is just too young with too many things as of yet untried, along with the fact that I doubt funding will dry up any time soon. (There are too many interests, from Nvidia wanting to sell more chips, to Microsoft wanting to sell more productivity, to defence, and political concerns between the US and China.)

Yes, it won't go on forever, but also this time seems qualitatively different from the past AI cycles. (Granted I was not alive then)

Well there goes the most solid security measure to ever exist.

No one's gonna hack floppy disks.

So I've been wondering...

Why not just have ISPs block IP addresses? Have IP blocking default to on, but allow the customer (who is over 18) be able to opt out.

Then have the services split up under-18 and 18+ IP addresses?

Seems much better than having to have websites magically determine age. It puts most of the onus on ISPs, but would be fairly simple on their end, and little to no affect on adults or parents who decide to take responsibility.

(I guess this is more in context of Louisiana law, not necessarily UK, which has more of a encryption aspect as well.)

What I've always kinda thought is that each congress person gets a large amount of budget to spend on staff. ~order of 10 million per congressman.

This would allow for the establishment of in house knowledge groups/ staffers, particularly with regards to policy, rather than having to rely on lobbiests. It would also reduce the churn from congressional staffer to industry lobbyist.

Or at least that's my thought. The us gov spends a ridiculous amount of money, spending more in the organization meant to monitor and control that spend. If you even have a half a percentage gain in efficiency would pay for itself.

This is forgetting the ability of congresspeople to be more responsive to constituents.

The real thing it needs is lower latency, imagine that demo without the 10 second gaps between speaking and response.

What is the use case for valgrind vs memory sanitizer?

So what I don't get is why they don't have it as two tiers, aka if you make software and sell it. Your software and all dependencies must be certified.

The people doing certification must have a active hand in the development process of the specific software component. who, at their option can charge for certification.

This would allow open source projects that are used in industry to charge for certification labels, to the commercial companies that require it. But non-commercial which does not need a certified stack, is business as usual.

Seems like it would incentivize funding of open source, right? As long as the fees are low enough, no one will fork. But since only the people maintaining the project can make (and optionally charge for) certifications, it would incentivize knowing your software stack, and paying for maintenance/contributing back.

Forget that, have all members of a meeting be ai, each person gives them talking points.

It will either break things, or the meeting was pointless in the first case.

Presumably they don't charge credit cards individually...

So coming from c++ these factories are sounding a lot like compilers, how does a framework differ from a new language and standard library?

Or for that matter, an os and POSIX?

A language is just a standardized framework, and languages are created to solve general cases of specialized problems.

Sure, I don't need the python framework to do something, but if I don't care overly much about performance, why would I want to deal with manual memory management of c.

And just because c exists means we should not use c++ or rust?

Choose the right tool for the job, but starting at a high level of abstraction and working down as needed often makes a lot more sense than starting low and working up.

Maybe don't use unproven frameworks and all frameworks has a different level of provenness. yes there are some people out there that don't need anything, and can program a http server and website in assembly. But it's either a toy or fun side project if they do.

Wait... There was only one month where I was working on "finish your thesis"

So, I did not realize these were all supposed to be the same location.

I find it funny that my best guess was what appeared to me to be a Ohio conference room.

I believe that performance degrades, as the models end up training on data that is more homogenized and less diverse when compared to original data.

At least according to this paper (I think, FYI I am not a expert) https://arxiv.org/abs/2305.17493

For what it's worth from limited testing, (one example) gpt3.5 does have enough knowledge to determine the why.

I was curious what was going on in a sine cosine lookup table and that was doing linear interpolation between the points.

Feeding it into chatgpt, added the appropriate comments for each step.

It was fairly obvious, but had taken me 10 minutes to track down and realize what was going on.

No way would I ever run a random exe.

I only run random bash scripts from the internet with sudo permissions

Parts of US have this, (where it snows and is dry enough in the winter to be an issue.)

You need to disable intelligence (there is a popup)

If they pay me at contract rates double my current salary, and in my particular case my current IP assignment clause/the problem/ law makes it ok. Sure, pay me money to do something like what you guys need done.

Probably most useful if it's a contribution to open source.

Clearly the most advanced model we've seen to date.

It's already trolling you.

I wonder how many companies will be doing more bugfixes now...

For example, the initial product generates "hello world".

What it was supposed to do was control a robot to automatically do pick and place.

It's Definitely a bug that the program failed to work, it's even tracked as a defect in the issue management. No R&D involved just fixing a pretty severe software bug, namely that the product does not work.

Why? What do you think the better model is?