Hacker News Reader

Comment by epistasis | Hacker News Reader

Right now AI is a lot like VR headsets. A really cool tech trick but not something that most people want to use regularly. Let the interested early adopter at it, but forcing this on to users will not go very well. It's just too creepy.

Silicon Valley got pretty lucky with figuring out social media, but I think it still does not understand most people very well.

View on HN

I disagree. ChatGPT reached 100 million MAUs 2 months after launch. It’s one of the fastest-growing consumer applications in history.

Anecdotally, lots of my non technical friends (and me) are using it for everything from cooking to learning a foreign language.

Lots of my technical friends are using it for side projects on the weekends. I’d say it’s the top new technology all of them are working with or incorporating into their workflows.

I and all of my teammates are using it to help us write sql and answer basic programming questions.

It’s clearly a way bigger deal than VR right now.

The problem here seems to be that Snap rammed this feature into their product in a really awkward fashion that doesn’t make sense for their users. Hence the backlash.

source: https://arstechnica.com/information-technology/2023/02/chatg...

You're inside a bubble though. The 100 MAU is certainly totally misleading, maybe 1/10 of that in reality, and something that's only out for a few months can easily rollercoaster up and down as people try it once for novelty and then forget it.

I have a good group of friends who keep me grounded - we all went to my average state school alma mater, and none of them are in tech.

Not one of them has brought it up, no one uses it or cares about it, and only two of them even know what it is beyond having seen some headlines.

The playoffs have gotten about 200 texts recently, AI 0. This is closer to the reality on the ground.

I'd venture almost all of the hype is students and kids who are excited to see what mischief it can help them achieve, techies who just like playing with new things, and companies trying to cash in. All of those are non-durable.

“You're inside a bubble though. The 100 MAU is certainly totally misleading, maybe 1/10 of that in reality, and something that's only out for a few months can easily rollercoaster up and down as people try it once for novelty and then forget it.”

What’a your basis for saying this is misleading and doubting that figure?

Anecdotal friend groups aside, if their was no user traction, they wouldn’t be getting a ten billion dollar investment from MSFT.

Their growth in web traffic is also pretty impressive:

In my personal and professional life I’ve been using it every day and happily pay $20 for premium. It has replaced google for me for a huge variety of queries.

Because they’ve had limits on accounts combined with trivial registration of new accounts, combined with counting enterprise “accounts” from stuff like bing or api usage which really is just the same people using them over and over. That and startups tend to look at numbers that please them once and never question them. I don’t doubt they had a ton of accounts sign up. But MAU after 2 months isn’t a statistic it’s a data point, and coming from the most unreliable source possible.

Incredible growth in traffic is easy to explain - a lot of big companies and the most influential investors invested a ton of money in this. They have connections to all the top media sources. They pushed this stuff absolutely everywhere. It was a massive media blitz. You’re being manipulated. They’re hyping AI, they’re talking about doom and fear, they’re getting front pages everywhere. Everyone wants in - media wants in on the hype, social media users realize they can gain tons of viral views, it’s a giant pit of self reinforcing hype. That happens with things. See Pokémon Go, and then what happened a short while later. Or the crypto bubble. Or any number of bubbles. You can’t base your predictions of future success on self published popularity numbers after a huge and expensive media blitz.

What's the source of the 100 million MAU number? I have seen it thrown around but haven't found something from OpenAI stating that.

AI is already way more important than VR. I have had a lot of people tell me that they are already making serious economically-impactful use of AI both personally and professionally, many others are the same.

The failure of Snapchat here is doing AI for the sake of AI.

My most successful use of AI so far was that Bard's answer was so useless that it made me realize the actual problem. Invoking Cunningham's Law isn't really a success on its own merit though.

What was the question and what did it answer? For me as a startup founder, it's cutting down my coding time to 1/4 of what it would have been, since I can ask ChatGPT clarifying questions and ask it for direct code examples.

It's closer to blockchain. Unprofitable companies are rushing to jam in something, anything in there so they have something to tell Wall St on the earnings call.

Companies were rushing to VR/metaverse as well. I think we can just accept that any new shiny piece of tech will attract an army of grifters trying to capitalize on it and those comparisons won't tell you too much about the tech itself. Fwiw I've seen a lot more LLM demos that made me go "this idea can probably work and be used by millions" than those for VR or blockchain.

I couldn't disagree more. It's clear AI is going to have huge impacts on the world. It's just getting started though.

Imagine every time you call support, instead of having some annoying rules-based automated system, you're having a natural conversation with an AI that is fluent in your language. That's possible with today's technology, it just hasn't been built out and deployed yet.

Imagine instead that you could speak to a human. One with the authority and skills to truly help you. That is the dream, not a less shitty robot.

Yeah I love companies that just put a real human on the phone who knows what they're talking about. All of the problems I'd want a support number for are impossible for an AI to solve, and putting a speaking AI in between me and a real human would just annoy the crap out of me.

I bet Comcast is working on it as we speak.

Tons of companies have already replaced their actual support chat with AI bots, it's absolutely annoying when you want to get ahold of someone. Some allow you to just ask to get ahold of a person, others don't allow it or have extremely limited hours.

That already wasn't the case without ChatGPT.

Even if AI has huge impacts on the world, and does a lot of great things, that in no way negates my point.

Right now, almost nobody have an interest in AI chat bots. Even fewer want to have it in SnapChat. And honestly I really don't want an AI bot for customer service, especially at the current stage of tech dev, because the AI bots are completely unreliable and hallucinate and lie regularly. (Which is to say that I heartily doubt your assertion that current AI tech would enable customer service bots that are useful. They produce wonderful prose, but semantics and action are not there, and we don't have training datasets yet to enforce accuracy in prose or action)

So let's say they become useful, great. But today they are still a VR headset, something that's fun to experience for a bit, but which I don't want to be part of my daily experience. It's a novelty, not a useful tool, for nearly everyone out there today.

Make it useful, attractive to most people, and we are in a different regime, and forcing features like this on all your users might be seen as a positive rather than a negative.

I think I’d rather have a clear set of options rather than having something that is going to essentially randomly, uncontrollably respond to what I input. What’s more annoying something that clearly can’t help you, allowing you to move on, or something that will endlessly try to appear helpful while still being manipulative in upselling and avoiding solving my actual problem?

It might be better in some ways, but I don't expect it to be perfect. The AI is likely going to have safety filters and not answer user questions even when they may be safe. Companies may also be incentivized to keep conversations short to keep their token count and therefore bills low. So users will likely not be able to have long conversations.

So did fascism. Not the best metric.

I don't know how you can think that. Random companies scrambling to get in on the hype to look good for investors is one thing. But the actual good applications coming out of AI are already getting integrated into people's daily workflows. ChatGPT is a genuinely useful product to tons of people.

Most people I know who bought a VR headset on the other hand thought it was cool for a couple weeks then it was never touched again.

Hard disagree. Anecdotally, everyone in my friend group uses chatGPT/Bing many times a day.

Reply on news.ycombinator.com