Hacker News new | past | comments | ask | show | jobs | submit login
Predict HN: In 10 years, AI partners are the norm
22 points by warent on Dec 5, 2022 | hide | past | favorite | 20 comments
It seems like many people feel two opposing ways about AI: either AI is "meh, kind of basic" (even ChatGPT or Dall-E!); or that they're so good it's becoming some kind of existential threat to jobs / careers, purpose, etc.

There was a time when flip-phones were peak consumer tech, then 10 years went by and we all suddenly had touch-screen pocket computers.

Well, right now we're in the "flip-phone" era of AI. In 10 years from now, mark this prediction: nearly everyone will have an AI counterpart that is trained and unique to each individual.

Nobody is being threatened; everyone's baseline is simply being leveled up. People who have no experience in a field will suddenly be enabled to operate in that field alone. Meanwhile, people who already had experience in that field will now be even more effective and efficient. Therefore, professionals still exist.

In other words, for skill levels, "level 100" becomes the new level 1, and "level 100,000" becomes the new level 100.

As an aside, internet search will become largely deprecated as our AI become conversational human interfaces to access all the information we need.




I don't think humans are going to 'partner' programs, but they will use them to reduce their dependence on people they don't like which means fewer interactions with people outside the emotionally meaningful ones.

But what i think is a bit worrying is that people will start enforcing AI standards to their relations with other people. We see that already when people enforce Internet standards (like cancelous attitudes) to their real-life relationships and it doesn't always end in more happiness.


Some people are already capable of rapidly gaining domain knowledge from openly available sources. Reading the top 10 books in a field, watching a 100 hours of recorded conference talks from expert practitioners, and reading about a thousand articles will give you more than enough context to operate in a domain, especially if you do some interviews with experts yourself and are DISCIPLINED about how you are learning the domain and building a mental model of the domain.

You could say that google is already halfway to being an AI assistant with how they try and infer the real intent behind a search query. The question is what is the cost when an AI is wrong and you don't know that they're wrong. I would not want to start a startup based on 'knowledge' from ChatGPT.


You won't start a startup based on knowledge from ChatGPT, you:

- will write a better performance review for your peers

- will find better deals on Amazon

- will sing custom karaoke songs (that mention you and your kids on that great match last day) on Saturday evenings

- will make better predictions when betting who's gonna win the next world cup

- will write more clickbait tweets

- etc.

All of this assuming the AI you can get access to does not get vandalized by Google, Apple, Facebook et. al and the first answers the AI is giving you are all sponsored by ads... which I certainly believe will happen. I guess AI assistants won't come with a Premium feature (without ads) just in the same way Google Search is not available without ads.


One thing is for sure. You can get lot of investments using keywords like AI for next 5 years or so. Lot of "AI Based" startups are going to popup and investors will line up to give them their money.


So we'll all get a daemon like in His Dark Materials?

You know what? I actually don't mind that timeline.


The first use case I could think of was a AI trained on my team's codebase. So that whenever we face an oncall incident or unexpected crash, we'd have this AI partner (that has been finetuned on all of our code / logs / previous bugs / documentation) that we could ask to help us debug.


I’m not sure about the everyone having a personal AI, but I do think this will render the google search type experience obsolete. Being able to ask complex questions and get instant information on demand will change the way we interact with the world


That’s the personalized, human-centric vision.

The big five aren’t going to give that to you. Are you going to train your LLM yourself? Personalize it? How’s it on servers in your house? No. You’re going to use the one they give you and it’s not personalized, it doesn’t work for you. It actively works against you, advertising to you (which is just convincing you to buy something you ordinarily wouldn’t) and otherwise extracting money from you. The reason companies are investing billions in AI is they expect a return on that investment. The return will be funded by money they extract from you.


If people keep talking to AI all the time, how will they recognize emotions and which ones are meaningful? If people used social media only to keep their connections with real people alive, it wouldn't be much of a problem but we know what happened there.

Looking at all the improvements in LLMs, I am of the opinion that they are worth it ONLY if they deliver some massive breakthroughs in physics, maths, biotech and comp.sci. The way it seems to be going, we'll be getting the parlor tricks, the SEO, the social cooling before it presumably becomes good enough to actually help humans.


Unlikely. We can't even get a decent home automation voice assistant today. Total garbage at this point. So bad Amazon is getting out of Alexa. If we have one of those in 5 years, we'll be lucky. People like Elon Musk are over promising on every front and there's going to be hell to pay because of it. We're going to see a reversal in investment and a general public turn away from this tech. A new winter is coming. Niche products will be created, but these will be amazing specialized tools. Another 50 years for AGI, if we make it that long.


That reasoning feels sort of circular though. The hype about ChatGPT is because that bot has (ostensibly) archieved exactly the kind of thing that Alexa, Siri, etc have always struggled with: Natural conversations.

If that progress is true, I think we can either expect Alexa and co to be upgraded with GPT-x tech at some point or a new competing voice assistant to emerge. I think it remains to be seen if Amazon will still be pulling out at that point m


Text interface is a totally different domain than human speech.


You assume AI will get a lot better. That isn't a given. AI has hit a plateau many times.


Feels like this prediction is unfalsifiable since we've had a personalisable conversational human interface AI in our pockets since the launch of Siri in 2011. Not sure there's much sign of it progressing beyond a fun but erratic UI though


Using the phone analogy, Siri is a string with two cups on either end.


Seems a bit optimistic to imagine the leap from the string and cups toy to universal iPhone in ten years, and I say that knowing that an iPhone is an actual thing...


Oh remember the chat bot craze of 2017/18?

OpenAI is great, I think lots of people will built on top of it or use it and it will increase the productivity of the human race.

Also, I think we will have AGI within 10 years.


So basically we will have AGI in 10 years?


AGI has always been 10 years away.


It has literally never been ten years away, until this year. Or for the vigilant, since GPT-3 was released.

When has it been called 10 years away before? Not since the 1960s.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: