I'd say that an LLM perhaps represents something that would make a machine slightly more conscious.

The human brain also calculates a response based on some input, it's just a lot more complex.

If we build a machine that is as complex as the human brain, then yes, I would say that this is consciousness. The fact that we are able to explain how it works should not matter.

If a human is a 100 on the consciousness scale, an LLM (with memory) is perhaps a 4 or something. The interesting question is how far on the scale do you have to be to have certain rights etc. This is something that people are already discussing in regards to animals, i.e. a dog has more rights than an ant.

The problem is that getting an AI to answer a question is trivial. If I wanted to know what an AI has to say about the topic, I would just ask myself. Sending AI output has, as the author writes, the same connotation as sending a LMGTFY link. It does not provide me any value at all, I know how to write a question to an AI, just as I know how to use Google.

They want to explore what is possible and what sticks with users.

The best way to do this is to just push it in their apps as many places as possible since 1. you get a nice list of real world problems to try and solve. 2. You have more pressure on devs to actually make something that works because it is going into production. 3. You get feedback from millions of users.

Also, by working heavily with their AI, they will discover areas that can be improved and thus make the AI itself better.

They don't care that it is annoying, unhelpful or uneconomical because the purpose is experimentation.

Also love the humble brag. "I've just closed my 12th bug" and later "12 was maximum number of bugs closed by one person"

I've seen the opposite problem many times. Someone has confidently implemented something terrible, and I'm just thinking, why didn't you ask anyone about this!

Posthogs solution seems to be to "only hire very good people". Which is kind of funny to me. That might be possible in a hip company working with exiting new tech. But for many companies such as "boring" companies with old outdated stacks or start ups with no money, that just does not seem to be possible, at least in my experience.

Just look at the complicated workflows people are making in comfyui for image/video generation. Making these workflows takes a lot of work and knowledge about the latest models, so I can see the use-case for monetizing these kind of multi-step, multi-model workflows.

Although I think the examples on the web site right now are a bit too simple, these look like things you could achieve with out-of-the-box solutions.

As the article illustrates, users will ignore all warnings and get themselves scammed, and so the last resort is to not even give them the option.