AI Will Be Like a Super Smart Person Who Knows Everything About Your Life | Hacker News Reader

AI Will Be Like a Super Smart Person Who Knows Everything About Your Life

futurism.com

15 pointsrobblack20 comments

> the best use of AI would be a "super-competent colleague that knows absolutely everything about my whole life, every email, every conversation I’ve ever had, but doesn’t feel like an extension."

Not if I can help it. That comes off like a threat.

Right, I already have a spouse. Unlike a random box of algorithms, our relationship creates a whole stack of mutual incentives so I know I can trust her output. She's not going to slowly start rebooking all my meetings to take place at Starbucks because they wrote her a check.

I'd use something like that, provided it lives in my closet next to a big red physical kill switch.

Those are both part of my minimum requirements to use something like this as well -- but I'd also want my security game to be a lot stricter. More strict than I would be willing to live under.

The accumulation of so much data in one place is an inherent security problem, even if that place is my closet. I want to segregate data on different machines/locations so that a successful attacker can't get access to the entire trove. If I can't do that, then I need an absurdly high level of security. The risk is just too high.

The possibility of that super-smart person being hacked ... is alarming.

For me it's not even "hacked" that's the problem, the issue is cloud, subscriptions and the intersection of government and corporate interests.

My phone is an amazing tool but it also enables quite pervasive surveillance, including realtime location tracking.

Will this new thing be a "super powered assistant" or a superficially helpful snoop who's going to sell my interests and attention to the highest bidder? While co-operating with surveillance dragnets on the side? Is there any possible way that it turns out NOT to do those other things given the incentives for all parties and current regulatory environment?

I think it won’t be long before open source models will be good enough to be good enough, such that the responsible thing to do (for those technically capable) will be to learn how to self-host them. Give it two, three years and I think there will be open models of GPT-4 ish quality that can be run on <$10k of hardware, customized and tuned based on whatever data you want them to be, and capable of agent-type behavior like searching the web. To have a tool like that self-hosted on a server that I control is what I’m waiting for.

just because you can doesn't mean it's convenient

you can theoretically run AOSP, but Google will do everything in its power to make life miserable for you. most people generally run Samsung, pixel, or Apple

It's bad enough that they live on somebody else's computer

The fact that any super-smart person can be 'hacked' is nothing new.

It's no different than how smart human beings can still be brainwashed/indoctrinated/programmed into believing anything (whether by 'the government' or malicious actors or manipulative sociopaths).

I've been using ChatGPT's new memory feature. It's buggy, but when it works it's quite cool. It feels like a glimpse of something big. Since it began making me aware that it remembers my activities, I've been pushing myself to achieve things. It's like having an audience. I wouldn't have expected that: when the machine is smart, I want to impress it, just like I'd want to impress a person I respect.

> I want to impress it, just like I'd want to impress a person I respect.

Why do you want to impress it? Why do you want to impress people you respect?

Non-stop, supercharged self-improvement is here.

It will be interesting to see how this vision of AI unfolds and how society navigates the implications of such advanced technology.

I mean, they sold us automation as this great thing that would bring the two day work week by 2000, solve poverty, solve malnutrition, &c. Now they're doing the same with """ai"""

If the undelying system is rotten whatever grows on top will be rotten too

I think actual real life is so complicated that we can't really guess what specific technological advancements will rocket us into silicon valley imagined utopia, we speculate but we don't have any way of accurately predicting this.

Will it be the self-driving car, humanoid robots, chat bots, Google search with adsense (tm)? Hard to say. Each new thing brings some positives and negatives but it really seems the bad has an uncanny ability to keep up with the good most of the time.

I can confidently predict that no technological advancement will bring us into that utopia. With our current level of tech, we could easily be there now. At this point we don't need engineering technology, we need social/political/economic reform.

You're likely correct :)

We have the tech for decades, we lack the social and political decisions

And controlled by megacorps and governments.

Wasnt there a whole genre of scifi warning us about this?

Reply on news.ycombinator.com