lpolovets

I'm surprised the introductory 50% discount is good for 4 months. It seems like frontier models release new versions every 2-3 months, so raising prices in 4 months seems like a bad plan: you're effectively planning to charge users twice as much for a model that is no longer frontier.

I think stuff like this is much better settled with compensation than legislation. For example when I was an engineer 15-20 years ago, my friends would deprioritize jobs with lots of on-call needs, but would still take the jobs if they were exceptional in some way or the comp was especially good. Why can't we do that in most other job categories as well?

I can think of a bunch of areas where this kind of bill would degrade people's experiences with businesses:

- handoffs will suck. Someone's shift ends at 2pm and you take over. At 2:05 you realize you need to ask them if a client issue got resolved or if they did some important item on their task list. But now you're not allowed to. Lots of time will get wasted as a result.

- scheduling will get slower. This would impact founders if it was in California! For example VC firms (like mine) have executive assistants, and the expectation is something like "work 7-8 hours a day, and check email a few times in the evening in case something pressing comes up." But if you can't ask your EA to do a quick email check a few times in the evening, the thing that could've been scheduled today for tomorrow will instead be scheduled tomorrow for the future.

- if businesses need to hire more just to have a little more around the clock coverage, then prices go up for everyone. E.g. if 5% of work happens unpredictably between 5pm and 10pm, then either that work can get ignored until the next day (previous points), or someone can be hired to work 5pm-10pm -- but eventually that extra cost gets passed down to customers.

Related to this, I hate how aggressively Google pushes Gemini and all of the privacy implications involved with that.

1) Lots of features got moved around and there are now many "Write with AI", "Generate image with AI", etc buttons polluting user interfaces even though I don't use them and don't want to use them.

2) Actually, I would use some of these features if I didn't have to do a full opt-in to Smart Features for Google Workspace. If I'm writing a blog post and want to generate a cat picture, that doesn't mean I want to turn on invasive AI-enhanced features in every Google App under the sun. Gemini's chat interface is similar from I can tell: either I can see my search history but Google can train off of it, or if I don't want Google to train off of my chats then I can turn History off but then I can't view it myself. Why isn't there an option for me to see my history but not Google?? They're just the worst at caring about UX.

It's sobering to think that if every single person in Malta -- an entire country! -- signed up for ChatGPT and used it weekly, ChatGPT’s WAU would increase by only a few tenths of a percent.

This is secondhand and not firsthand knowledge, but one thing I've heard from a number of agtech founders (and sometimes farmers when we do diligence calls with them) is that hiring employees is very hard because of seasonality. E.g. you might need 100 workers most of the year, but 400 for a few weeks when it's picking season. And of course if you are in an area known for a single crop, all of the other local farms have the same seasonal labor need.

For these types of businesses, paying less for labor is a secondary goal and just having enough labor (human or machine) is the primary goal.

This feels premature -- we've only had great AI capabilities for a little while, and jobs are not replaced overnight.

This reminds me of some early stage startup pitches. During a pitch, I might ask: "what do you think about competitor XYZ?" And sometimes the answer is "we don't think highly of them, we have never even seen them in a single deal we've competed for!" But that's almost a statistical tautology: if you both have .001% market share and you're doubling or tripling annually, the chance that you're going to compete for the same customers is tiny. That doesn't mean you can just dismiss that competitor. Same thing with the article above dismissing AI as a threat to jobs so quickly.

To give a concrete example of a job disappearing: I run a small deep tech VC fund. When I raised the fund in early '24, my plan was to hire one investor and one researcher. I hired a great investor, but given all of the AI progress I'm now 80% sure I won't hire a researcher. ChatGPT is good enough for research. I might end up adding a different role in the near future, but this is a research job that likely disappeared because of AI.