Ask HN: A New Decade. Any Predictions?

Because car-free city is the future and of 3 technology leaders only USA will fail to develop sufficiently good public transport in the next decade. It is the only culture so deeply attached to cars and individual houses that it will be always behind others in urbanization.

1. Deep-water offshore wind power will be an established technology, and capacity deployed will be >50% of shallow-water offshore

2. Addition of 20% low-carbon hydrogen to natural gas grids will be common in Europe

3. Still no commodity optical computing, still no restoration of full-blooded Moore's Law, CPU core counts will be huge, we programmers will still be rubbish at effectively making use of them all

4. Rust's lifetimes prove to be the crack in the dam for type systems which regulate more aspects of program correctness in widely-used production languages (as widely used as, say, Erlang is now)

5. One of the 2028 Democratic presidential candidates will be openly furry

Here are my predictions:

1. Countries and private companies will launch stablecoins and it'll be the most used payment medium in developed countries.

2. Bitcoin will grow, but it'll not be the biggest cryptocurrency anymore.

3. Total market cap of all cryptocurrencies (including stablecoins) will be ~$8T.

4. Around 50% of the software jobs will be remote.

5. USA and China tradewar will turn into something more scary and people will be increasingly more wary of a third world war.

6. Apple releases their first iteration of AR glasses and it'll use an iPhone as it's driver.

7. Renaissance of piracy through decentralized systems like Filecoin. We'll have illegal products that will be easier to use than Netflix, and more content. Netflix, Spotify, et al. will never be able to have a complete library of all content, but it's pirate competitors will.

8. OpenAI will start building products or provide services for consumers or businesses. They'll be very profitable and have a valuation of around $30B.

I have some background in speech recognition, translation, machine learning, so many of my predictions will center around these topics.

1. More consumer robots, more fancy variants of cleaner robots.

2. Autonomous cars will be there, at least simple variants which work under some constraints (e.g. highway). (I think this is level 3 or 4, not sure how these levels are defined.)

3. More world-wide conflicts, because of politics, and also because of climate change. I'm optimistic that a world war will not happen, but it will become worse and worse.

4. The deep learning hype grows rate slows down a bit. Custom hardware becomes more important, as we are currently hitting the ceiling with computing power, which will not grow as much exponentially as it has before. Custom hardware will first focus on making existing models cheaper and faster to run (think of TPU), but at some later point also lead to different models and learning algorithms (Neuromorphic chips).

5. Speech recognition is a solved thing, more or less. The focus will shift towards embedding a speech recognition system/model into some bigger system, like an assistant, i.e. with parts like natural language understanding/processing, dialog systems, etc, and that will be more like an end-to-end system, not so much separated anymore. I.e. one big neural network for everything. Maybe in the beginning parts of it (like the speech part) are trained individually, or in a pretraining scheme. Current heuristics and hardcoded rules will go away and be replaced by neural solutions.

6. Meta reinforcement learning will become more important, and bigger more general neural networks. Generality means that they will solve multiple tasks at once, similar to the dialog system I outlined, but this trend will also follow in other domains. It will all go into the direction of AGI (which basically is exactly that: an AI which is (more) general), but not reach it fully yet. Meta RL will also go along side with online/continual learning, and being able to learn new things at inference time. There will be stable basic models / learning algorithms which can be easily plugged into some system as a building block, and it can be treated like a black box and will mostly work (think about AlphaZero or MuZero but more advanced, more general, almost no hyper params to tune, all automatically). Meta RL will also solve sample efficiency to some degree.

7. Much more electric cars, maybe being the majority.

8. Not much people will care about Bitcoin anymore. Maybe other cryptocurrencies though, but not sure. Proof of work is not a long-term solution.

9. China will dominate in many aspects, such as economy, but also research.

You had me up until here: 'Web development matures and a "standard" stack is accepted, all in JS.'

The Node ecosystem is a mess and it starts with core being a hodgepodge of deprecated mush. I think it's more likely that wasm picks up steam.

Rails, PHP, Django, etc aren't going anywhere.