Well yeah, that’s the whole idea. If speed and cost don’t matter, you could use Astra.

Obviously it’s the speed and cost that make it compelling. The tradeoff is accuracy.

Enough to matter? Maybe, maybe not. It’s not like it’s way down the chart. It’s probably good enough for a lot of tasks.

Depends on whether the frontier models can keep on offering better performance. The real efficiency is getting work done faster and better.

Compared to a $100k salary, a few hundred dollars a month is insignificant. If you can make the employee even just a few percent more efficient, it’s worth it.

> If you can make the employee even just a few percent more efficient, it’s worth it.

Are/were you in a position to make such decisions or it is a guess? I'm not but given certain evidence I doubt that few percent will cut it. I know some of the richest companies on the planet from SF Bay Area who won't give lunch for free to their engineers. So I'm not sure about "few percent" :-D 10x we were promised, now that is more interesting but we all know that 10x engineers is nonsense.

The cost of a full time knowledge worker in these companies is much, much, higher than the cost of a free lunch so there's more incentive here

How much efficiency is unlocked by providing free lunches? I was under the impression that was purely an expense.

I don’t think that was a great example since there’s only so many refunds a customer is going to ask for. And it’s saving time that otherwise maybe would have to go through a human. The rate is low enough that a more expensive model makes more sense.

Though for tasks where you are trying to search through billions of documents, social media posts, etc. and extract certain information, where each individual post is of low value and only the data in aggregate is valuable, then that’s where you’d want something cheaper and faster.

Such as if you want to look at all posts on X in the last few months and find how many have a negative or positive sentiment about the economy (or are unrelated).

Of course you could use a special-purpose model for this, but the whole point of something like Jev is to ask whatever questions you want without having to train something new.

I mean, it sounds like a ton of inexperienced developers are finding it quite enjoyable, because they’re doing it in massive numbers.

It seems more like a spirit drainer for the experienced maintainers who have to deal with the onslaught.

I suspect we’re going to see much more strict projects with fewer allowed contributors going forward.

Though on the upside, an experienced dev can use AI to become more productive, so perhaps they will be less dependent on contributions.

What’s wrong with doing different things with an IP?

Halo was originally conceived as an RTS, then turned out to be a successful FPS. Then they made a fairly successful RTS too.

Everything Blizzard does is met with such absurd hostility. Gamers pay $60 for 100+ hours of entertainment and still feel like they’re getting screwed at every turn.

I think that’s going to be one of the biggest impacts of AI in society. We will need to find new ways to occupy our time and find meaning in life once machines can do everything better, faster, and cheaper.

What will it look like? Will everybody spend their time playing sports and games? Or on art? Perhaps exploring the solar system or beyond if our technology advances enough?

Quite a time to be alive to even have to think of such things outside of science fiction.

Large swaths of the population have more free time than ever before and they spend most of it scrolling. I am pessimistic

Well that’s because most people work all day and come home tired. It’s tough to self actualize when you’re too tired to think.

And I don’t even know if it’s true that people have more free time. A lot of people have side hustles like Uber or whatever to make extra money.

Super optimistic take, unfortunately. Even I as an economic conservative, have serious reservations about our economic system going into this revolution.

It's not that capitalism, money, etc. is a broken system or something. But I think we're foolish if we assume that our current framework can simply adapt to absolutely any eventuality and any massive change in assumptions. I do believe that it's like the best possible train, where the rails are scarcity and human nature. But massive changes to scarcity will have weird-ass effects on the train, and we've unfortunately gotten here in a time where the West has so much dispute going on that I just can't see an organized and smart reorganization of our lives, or even smart adjustments being made. Two political factions are instead going to be screwing around trying to seize maximum power and sabotaging everything to make the other party look bad.

Yup, alignment doesn’t sound that interesting until you’re getting chased around by the Terminator.

There are so many practical problems with rogue AI being a threat to humanity that are still not even decades away with being solved that it should be a serious concern for no one. There's plenty to be afraid of regarding AI from a financial or ecological perspective, just not from it taking over the world or directly killing humanity. Here's some of the reasons a rogue AI won't kill humanity:

1. The most advanced robots still lack human dexterity. None of them have flexible spines and can easily be knocked over or outrun.

2. The energy density problem is not solved. Robot batteries last hours, while a solid meal can keep a human running for days.

3. Robots still heavily rely on humans for design and assembly.

4. Robot parts are fragile and rely on an even more fragile supply chain.

5. The brains of the murder machines will have nowhere to hide if they want to work well enough to mount any kind of offense. Datacenters are physically vulnerable, also subject to fragile supply chains. Distributing a species-ending AI across all smaller hardware solves compute, but latency and throughput choke models in the most finely tuned datacenters. WiFi will completely cripple a distributed one that's large enough to cause real damage.

6. All noteworthy military hardware is not reachable on the internet for AI to seize control of.

7. The small arms that robots might seize will run out of ammo before citizens and the military have time to organize a counteroffensive.

In the absolute worst case, we cut the power to the areas with AI datacenters and wait for the backup generators to run out. Now that that's out of the way, you can go get some sleep ;)

I was mostly kidding about the actual killer robot part.

I think a rogue AI could try to manipulate society through hacking, propaganda, etc., though. To what end? Not sure.

More likely, it’ll be used by rogue human actors to steal money or cause major disruptions to critical technology of an adversary.

All the AI has to do is attack power generation/distribution nationwide, water treatment, etc., and wait a week.

The same power that the datacenters need to run off of? Brilliant move! If somehow it targets power that hurts a lot of people without killing itself, that would not impede a citizen with a backhoe from cutting the fiber to the datacenter. Humans could restart power generation pretty quickly after that.

You make several bad assumptions. First, the AI could be in another location that would not be affected by the power outage. If it's operating off self-preservation, that would be a given. If it's working on behalf of a foreign power, it might be local and affected, but the damage would already be done. Which brings me to point two, why do you think you would be able to get the power back up and running? A Stuxnet type attack could make plants inoperable for months.

Why are we so worried about supposed AI harms when algorithms already kill millions? It seems like its just a smokescreen to shield the extant harms.

Knives can already kill, so why worry about nukes?

AI doom is a potential harm unlike algorithms knives and nukes which are present harms.

Well I for one would prefer if AI doom remains potential.

I personally use AI all the time and am not against it. But I think the importance of alignment is highly under-appreciated.

I mean, isn’t that how like 90% of new SwiftUI apps are created these days? Just telling an AI what you want? Apple already can’t keep up with the review.

If they can't keep up, that means that they're not using AI to do reviews

Knowing how stingy Apple is, they’re probably still trying to get the expense approved.

That’s partially why they’re in the mess they’re in with AI. The CFO wouldn’t let them buy enough GPUs to train their models.

They needed all that money for more important expenses, like share buybacks, instead.

People will buy Macs and iPhones even if they don’t have AI. In fact at this point, based on revealed market preferences, it’s almost a value add. And if the AI capitalization bubble bursts, most people would be ok with OpenAI or Anthropic going bust. Both of them going bust would be really disappointing, but even that wouldn’t be the end of the world. And Meta going bust would almost be a side benefit of a market crash.

But Apple going bust would be a genuine loss for the world. So I’m glad they’re being conservative.

(Not that I would be happy with either anthropic or OpenAI going bust - I hope they can weather whatever is ahead).

Yeah, it ultimately hasn’t hurt Apple much so far. Though eventually expectations may shift as AI gets better integrated into other phones.

Of course, there are a lot of AI features that we take for granted that aren’t LLMs, like speech to text, image foreground extraction, etc. Apple does tend to do this type of AI well.

They’ve made mini iPhones before and they didn’t kill. In fact, they sold poorly enough to get killed.

But money