Ask HN: Tech Forecasts for 2024 – Any Surprises on the Horizon? | Hacker News Reader

Ask HN: Tech Forecasts for 2024 – Any Surprises on the Horizon?

36 pointssandwichbop26 comments

Curious to hear about your unconventional tech predictions for the next year. It's challenging to keep track of every innovative trend, especially with the potential for unexpected breakthroughs. My focus has been on AI, particularly since 2018, and I was amazed by the rapid advancements in transformer technology. I kept telling everyone I knew that this was going to be huge, but it felt like the boy who cried wolf until it actually happened. What insights do you have for those willing to heed a 'cry for wolf' in the tech landscape? What do you think will be the next big thing in 2024?

View on HN

Quantized Inertia is proven when the Barry-1 satellite[1] successfully shifts its orbit using a purely electric thruster (i.e. no propellant)

The laws of physics get re-written a bit, and Mike McCulloch starts looking like he'll get a Nobel prize out of it.

[1] https://db.satnogs.org/satellite/QDDY-8878-5291-1819-3935#ma...

This seems really cool. Am I correct in reading that this explains away the need for (some? all?) dark matter?

"Quantized Inertia further predicts that for objects with very low accelerations, such as stars at the edges of galaxies, the Rindler horizon moves so far back that it is close to the cosmic horizon so the waves of Unruh radiation are now damped equally all around, and the mechanism of inertia collapses. This explains why stars at the edge of galaxies can orbit faster than expected but still remain bound to the galaxy. As a result, Quantised Inertia predicts galaxy rotation perfectly without the need for dark matter or any adjustment (McCulloch, 2012, 2017)."

Source: https://quantizedinertia.com/about/

Yes that is why I asked. It wasn't clear to me if it was just dm related to galaxy rotation or if, by eliminating galaxy rotation issues, all need for dm goes away

People will accuse AI of interfering with the USA presidential election.

Someone will try to marry an AI.

Something more obnoxious NFTs will be a huge fad, then collapse.

Someone will try to convince people that their AI can predict the stock market.

Somewhere, an author of an image generator will be arrested for "creating CSAM" after the model is used by someone else for that purpose.

Most devs, including myself, will continue to work boring dev jobs while wondering when their job will be replaced by a combination of AI and low cost overseas labor.

Convincing AI with a human voice and evil script is used for scamming, at-scale.

Conversational AI robocalls.

AI finds its way into children's pets--a next-level tamagotchi.

A company releases AI-powered talking pets for the elderly/those with alzheimers. It seems both terribly kind and cruel.

Some terribly formed AI regulation passes.

AI can create whole movies from scripts, and scripts from suggestions.

LLM-based AI's get better at chess.

AI assistant's start to get decent.

"and scripts from suggestions."

Can't they do that already?

Sorry, didn't mean to imply they can't--I just meant it could be "suggestion -> movie" which is nuts to consider.

Some random predictions:

- OpenAI outages will cause more companies to explore self-hostable models

- US public education buys into AI tutoring or other services. A teacher will say "just ask your AI..." to their student.

- Meta's Metaverse becomes interesting instead of cringe

- People still remain confused on whether to call it X or Twitter

Your first point is interesting. I’ve been meaning to explore some of these options lately (llama.cpp). I’m a bit too smooth brained to build a good business offering around that.

Security predictions for next year:

- Continued information operations by nation states using social media, trying to influence US elections, next year around this time (regardless of outcome) about 40% of the US population will believe the election was rigged

- People will try to shoehorn AI into security ops, with little success due to not having great data

- CISO positions will increasingly become more hostile to accepting organizational risks, because of FTC civil complaint on solarwinds CISO

- there will still be new CVEs

Apple defaults to OpenAI as the default search engine.

A weak form of AGI that while not skynet does a passible job at most tasks.

That seems less likely than rapidly self improving AI, or AI that just doesn't work too good.

Why do you think AI will be able to do most tasks, but not improve itself? Unless you have a different understanding of "most tasks" than I do.

Our current method for making AI is largely a dataset creation problem.

An AI can’t just make meaningful data out of nowhere. It needs to have seen that data before.

As it stands, it can’t improve itself since it’s ability is restricted to what data it has seen.

I do think it’ll improve itself. Just not singularity like fast.

A bit like a child learning doesn’t make it an Einstein overnigh

Eventually it’ll pick up pace but not in 2024

Quicker than expected, there may come a time that for most people, not a single human has touched, or even seen their food anywhere between growing & when it ends up on their plate.

AI controlled farming + harvesting. Factory machinery to sort / wash / process / package foodstuffs, where the only humans are a few technicians. Fully-automated distribution centers. Self-driving trucks shuttling goods between those. And a 'housemaid' robot that arranges people's shopping & cooks their meals to perfection.

We're not there - yet. But with our existing globalized & industrialized food supply chains, and a bit of 'AI everywhere' sprinkled in, it's not hard to imagine.

- almost all AI startups based on OpenAI will fail (they have no defensibility)

- economic activity of LLM-powered self-governing agents seen in the wild

- copyright reforms lobbied by OpenAI and similar companies

What do you mean by "defensibility"?

Color eink displays integrated in surprising places will grow a lot. I think we may even finally see eink intersections where the lanes change dynamically depending on the light cycle.

I also think both eink and oled will make its way into fashion among a subculture of younger people, people who live wear, avatars. I can see a big opening for a social marketplace for light rythms and patterns that you can download into your jacket or hat. I imagine some people will pay thousands of dolars for dynamically changing unique "2nd skins" made by botique designers and sold via tiktok.

Bitcoin will reach $100,000!!!

HODL! Never stop gambling!

openai will find a reasonable model path between establishing and accumulating its own competitive barriers and building a win-win and mutually beneficial ecosystem with more entrepreneurs and companies. Rather than easily breaking through the projects mounted on it,

I’m waiting for the first AI enabled game. That could be a huge breakthrough in RPG gaming if done right. Just think about the mmo npcs you could have with decent ai

Rewind.ai or Adept.ai will have monopoly over the AI application-layer market

Reply on news.ycombinator.com