timmg

I guess it will be interesting if, in a week or two, OpenAI launches a "Fable class" model and it isn't blocked by the government.

There are a lot of things like this.

My favorite is how elegant solutions often look simple in retrospect. So if you noodle on a problem for a while and then come up with a clever solution: once you explain it to someone they'll be like, "yeah, of course."

Meanwhile the guy next to you that overcomplicates the problem ends up getting kudos for building something so difficult :D

I wasn’t assuming anything. I was asking whether the problem was bias — which we already see in some things that are highly regulated — or just wrong bias.

I’m trying to understand what people think we should correct for.

Ironically, in the US it is ok to charge men more for car insurance, since they cost more in aggregate. It is illegal to charge women more for health insurance even though they cost more in aggregate.

> sing an LLM to make decisions about extending credit to offer worse terms (say) to women.

In general, or if it isn't the correct answer?

Like: young men pay more for car insurance than young women (today). This is based on statistical models. Should they be outlawed? I think that is a very interesting question (but they aren't, today).

If the LLM was in charge, would it be wrong for it to charge young men more? Should we train that "bias" out? Or should we only train out biases that are wrong? And would that be different than how we train them today?

I don't know the answer. But I think it is less obvious than some people seem to think.

> There has been plenty of research that shows LLMs encode social biases.

At the risk of stepping into a hornets nest: is that different than "knowledge"?

Or maybe, what would it mean if an LLM had no social biases? (Would we ever agree that was the case?)

The way I've been thinking about it: there is too much money trying to pour into the market. That's why valuations are so high.

Maybe getting more of these big private companies public will bring valuations down a bit.

(Just my impression. No math or financial studies behind it :)

My first ever programming interview was like a group interview. There were three or four programmers asking me questions, one at a time.

The only one I remember was to check if two strings were equal (in C). I wrote (maybe buggy) code to iterate both pointers, comparing while looking for the null terminator.

The interviewer stopped me and said, “You should compare their lengths first. If they are different, you can exit early.”

I was pretty young and didn’t know much, but I explained, “But you have to look for the terminator to find length so it’ll take twice as long.”

He snapped, “There are optimized functions for that!”

I assumed he was right. Needless to say, I didn’t get the job.

Maaaany years later, I realized the std library was probably open source. So I checked (one). It was nice to be vindicated :D

There will be cheaper solutions. And they will generally be less capable than the more expensive ones. Just like most other products.

But my guess is that the cost of SWEs themselves mean that the more expensive ones will be worth the delta to most companies.

But time will tell.

> Pmf is this weirdly defined thing where "if you're not sure you have it then you don't".

I'm not sure if this runs counter to your point or not, but: I don't see any future where LLMs aren't a core part of Software Engineering. The horse is out of the barn. There is no going back.

I wonder if instead (or in parallel), Norway should build a set of training data and share it (for free) with all the model builders.

Seems like making the frontier models know Norwegian and their culture is a better (or additional!) way to reach the end they are going for here.

All of these concerns are valid. But none of them are unique to datacenters.

A golf course uses a lot of water. A factory can use a lot of power -- and generate pollution. A chemical factory could have all kinds of externalities (if not properly managed.) Heck, switching to electric heat (over gas) or electric cars over ICE for an area will also drive up power usage.

But we don't freak out when someone builds a golf course or a factory or switch to electric.

We have rules about all those things. Sound is one: you need to be within reasonable limits. Electricity usage is another: power operators always need to manage their load and expand generation (that's why we keep adding solar and wind everywhere.) Air pollution is similarly managed.

I can understand if people are concerned about "infrasound" -- why not pass a law that regulates it -- like other noise limits?

Datacenters may have specific potential issues. But none of them are unique to datacenters. And we've been managing these issues for hundreds of years.

Obviously, it's fine to be wary of any development in your area. But it seems like there is a certain amount of irrational(?) fear of datacenters. And I really don't understand it.

I saw a poll recently that people would rather live near a nuclear power plan than a datacenter. That's... their choice, of course, but doesn't seem logical to me.

I have heard several "concern stories" about them on NPR recently. Maybe there is a political component to it. But I do worry there is some kind of manipulation being done.

Also see: Failsafe. An earlier film.

Both seem to be influences of War Games.

Of course. (Though I didn’t see a mention of data centers in that doc? I did see “equipment manufacturers” mentioned specifically.)

If you click on the edit icon, you can see more data. It show 6 months worth. I thought it was longer. My mistake. But I guess I would still expect to see something over the past 6 months given the hysteria?

There’s another chart on that page that shows regional demand for electricity. That seems to have flattened or dropped. Not sure how to explain that, in the context of data center demand.

Maybe I’m reading something wrong. Or maybe there is an anticipated increase in demand?

It's not clear from the article whether this is just datacenters -- or if that is just the convenient boogeyman.

The grid operator for the northeast, according to my Governor, has been well-behind in building out infrastructure. Of course new datacenters cause more load. But so do new houses (we're building as many as we can) and electric cars, etc.

Not the person you are replying to: but I think that's the point.

Honestly, I hate "free tiers" for SAAS products like this.

For about 5 minutes, Google had a pay-as-you-go service on GCP for git. I used to use that, because I wanted to own my stuff. But (I guess) because everyone used "free" github -- and like a lot of other Google services -- they sunsetted it.

So now I'm using github for free. But would rather pay for my storage an usage with a (big) cloud provider.

Maybe it's just me, but I'd love a "ThinWarp" -- just the terminal with the great UI, etc.

I can run Claude Code there or whatever. But I personally don't need the AI in the terminal itself.

That cost has a lot to do with amortizing the construction costs of the plant. I expect that just running a plant is a lot cheaper than that.

This is about *not* decommissioning working plants.

> It’s incredible how forgiving you guys are with Anthropic and their errors.

Ironically, I was thinking the exact opposite. This is bleeding edge stuff and they keep pushing new models and new features. I would expect issues.

I was surprised at how much complaining there is -- especially coming from people who have probably built and launched a lot of stuff and know how easy it is to make mistakes.

Oh, I see.

Though, in fairness, that's probably the important part. Like a base model plus "coding smarts" is probably perfect for the situation.

But maybe not as much value as I was thinking.

I thought Cursor has started making their own models. Did I confuse them with someone else?

Not the person you were replying to, but: did you read that letter?

It is horribly argued. It's mostly poor analogies and non-sequiturs. It's no wonder Buzzfeed was the only place they could get to publish it.

This tweet shows it as a percentage of US GDP:

https://x.com/paulg/status/2045120274551423142

Makes it a little less dramatic. But also shows what a big **'n deal the railroads were!

> it seems like you are simultaneously arguing for a science that holds itself outside public opinion, and one that is beholden to it.

Apologies if I did a bad job explaining my opinion. But I was attempting to argue the exact opposite of that.

My view is that science should be the search for truth. And that if the truth is inconvenient for some political (or other) reason, so bet it. The truth is the goal. Full stop.

My feeling is that if scientists stop pursuing truth in cases where it doesn't fit their politics, they will (rightly, IMHO) lose the trust of the public. (Of course, in particular, those in the public who have different politics.)

As I said to another person on this thread: if scientists let their political views override their pursuit of truth, the public will (rightly) lose faith in science.

So when you tell them to "trust the science" -- be it vaccinations, climate change or something else -- they have no reason to trust that science.

I understand what you are saying and I don't disagree with the idea that bad actors will use science in bad ways.

But I think going down this path of denying (or hiding) science that can be used for bad ideas ends up causing (rightly, imho) a distrust of science -- which is far worse.

A distrust of science (not saying it was caused by this particular issue) is how we ended up with so much anti-vax sentiment in the US. And that is the reason we are seeing outbreaks of diseases that used to be minimal.

I think if you want people to "trust the science", you have to trust the people.