Ask HN: What are some reasonable black swan tech predictions for 2023?
One 1 hand we have "AI", promising chatbots that automate trival tasks, to the industry going back to old boring frameworks.
What are some tech society predictions you have?
What are some tech society predictions you have?
Peaceful protestors, as is tradition, were largely ignored. Acts of civil disobedience were ignored. Acts of obstruction are being frowned upon and the people involved are turned into villains by the media. There are people out there who went through all these efforts, faced rejection or deafening silence at each step of the way and are now out there further radicalizing themselves. Policymakers are still largely sleeping on the issue. This summer broke records and hit concerning extremes all over the globe. It feels like it's just a matter of time before something boils over.
If the USA keeps sending its construction crews to sweltering jobsites with no legal protection for their safety in the heat, hotter summer after hotter summer, people are going to suffer and die for no good reason and there will be follow-on effects. It is a heavily unionized industry that the entire country relies on and generally undervalues.
[1] https://www.washingtonpost.com/climate-environment/2023/06/1...
Black swans are what you can't predict with what you already know.
I think there's an important difference between "highly unlikely" and "black swan" — winning the lottery is the former, for any given number choices.
Waking up one day to the news that cow moos have been translated into English, they had an average IQ all along, and therefore all beef and dairy products are banned worldwide… would be the latter — it's not physically impossible given what (little) we know, but there's also absolutely no reason to suspect it a priori (unless you're vegan or Hindu, or…)
Given the OP's question, it's not clear they are using Taleb's definition, either. As GP pointed out, Taleb's black swans are by definition unpredictable and un-reasonable.
As such, you can use counter factual planning to prepare for some of these events. Interest rate risks would fit here.
This is complicated by his insistence that covid was not a black swan. As that seems to still fit the definition. That a pandemic could happen doesn't make it less impactful when it does. Would be like the expected earthquake on a fault line. Not knowing when it will come is a salient fact.
(That said: cow brains seem to be more complex than GPT-4 and LaMDA, and the latter managed to mimic sentience well enough to convince Blake Lemoine; a cow brain wired up like even a relatively mediocre LLM would certainly… be a surprise).
Black Swans are unpredictable events, so we wake up to Black Swan news.
On the point of cows having an IQ, all animals have an IQ, its just they cant vocalise our sounds well so whilst they can detect our sounds and respond, most cant even sign Makaton, so maybe we are the dumbs one for not coming up with a form of communication where they can communicate easily with us whilst accounting for their limited physical attributes and motor skills.
Basically, a large chunk of the unprofitable startups that raised in 2021 are going to fail at becoming profitable and be forced into fire sales or bankruptcy.
A recent example is Clutter, a moving company that raised $300m, who was forced into a fire sale for $30m.
These founders are going to learn the hard way one of the reasons why VC’s were willing to assign sky high valuations (bc they have downside protection)
Who is making money in these situations?
Is there a list of these firms
How do fire sales work
Sorry so many questions
One individual will be able to run a dozen in parallel, connected to the internet.
1 month later the first pixar scale movie is released, made from a guys basement.
Hollywood industry collapses
In general, first there was simply (weight * value) . Then people experimented with a trainable parameter that would select a set of weights for a given input. Transformers essentially turned this 1 dimentional discrete selection into a n dimentional multiplication, where the weights get modified by a product of values and a trainable parameter, i.e (weight * value) * value
To get further, 2 things need to happen. First, there needs to be investigation into being able to learn the network architecture itself. For example, what if the transformer learned parameters were instead computed at inference time? I.e ( ( weight * value ) * value ) * value. What if this was abstracted to some form where there was a learned parameters on where to get those values from?
Secondly, there needs to be some differentiable method of storing data between inference runs, so instead of input->output, you have input+memory -> output.
I give it another year or two, RunwayML's models are good but not as good as they could be yet.
Zero-determinant strategies applied to call handling. You want to cancel your subscription for a crappy thing? Your conversational tactics and state of mind during the call can be mapped to a large matrix and solved for you not being able to cancel and not having any legal recourse and thinking the outcome was your own idea.
*(Edit: mostly no longer on twitter)
ARM processors will become a common choice for Windows laptops and mini-PCs. AMD/Intel will occupy the high end and legacy spaces.
Smartphone makers will add LLM AI assistants to their latest phones to juice sales.
One of the full self driving companies will have a vehicle in a serious accident.
It's something that will likely get you laughed at by domain experts and the general public until it happens.
Using that as a framework, I'll predict something that I don't actually belive :D
> Aliens are real, they're here on earth and they've got extremely advanced tech. Some whistle-blower will blow the lid on the whole thing as part of a October surprise in 2024
That sounds like insane magical thinking and any reasonable person can easily discount it. But after it happened (by nature of it actually being real), it would have a rational explanation with telltale signs all along.
Not 2023, but I expect in 2024 or 2025 the current decline in security service spending by companies to bite them in the arse. See the first prediction.
1. "Choose boring technology"
The new situation for tech investments will force a change in how startups choose their technologies. The number of "innovation tokens" will be limited, as the runway is shortened and they can't afford to spend time on immature technology or pretend they have to design a product for Google scale.
Startups will be forced to make safer choices. Example: If you are an average startup, you don't get to choose an esoteric programming language or create your own database server, unless that language or database is your actual product.
2. "Low/No code is still not ready, but many startups will fail because of it"
Due to #1 many managers and VCs will consider lowering costs by trying low/no code tools. They may get an initial MVP quickly, but will waste lots of time and money once they move outside the market fit for their chosen tool.
4. "AI is the new blockchain"
We will see an increase in startups using "AI" as a keyword for attracting VCs. Those that focus too much on AI will often fail to provide a decent product, while those that considers it another tool in the toolbox will have an advantage. Prices on graphics cards will still be high, as the AI startups will need a lot of processing power.
5. "Sh*itification of streaming services"
TV/Movies are too fragmented and it will start hurting the streaming services. Consumers will increasingly become periodic subscribers, where they subscribe for a period and then pause the subscription until the next season of a good show is released. This will cause panic in the boardrooms and they will make many bad decisions that makes the product worse.
We will look at the years prior to 2025 as the golden age of streaming, until the product eventually becomes so bad that piracy and bankruptcies forces a more distributed alternative where a tv series is not tightly coupled with a streaming service (it may take several years)
6. "Twitter / X will still be there"
I think that they will be profitable and still be a major source of traffic. Some of the changes may be reversed (such as the name change and re-hiring moderators), but I think many in tech are underestimating the advantage they have by their size. They can expand their product and increase their revenue enough to become profitable even if they don't conquer the world.
I think all of us have some sort of fantasy in which we invent something which changes the world, makes us famous for generations to come, or something like that.
I always noodle on something like this. some new technology which makes the carbon in the air an asset to humanity.
But, of course, I know nothing about physics, material science, energy, or chemistry… and so it will remain a fantasy for me!
That said, I'm hopeful we could use cheap energy (e.g. the sun) to do something useful with carbon in the air, like plants do. See Running Tide as an example, they're trying to sequester carbon in a completely hands-off way using no added energy input other than the sun.
i.e. despite being behind and sanctioned they'll find a way to rapidly catch up and surpass the west
It's called available cash (roughly 3 trillion in reserves) and a self-supporting market that's 4 times the size of the US. And a government that has the impetus to 'get it done, and quickly'.
The West seems to delight in always shooting itself in the foot. (Though it's more like 'shooting yourself in the head' in this case.)
The West seems to delight in things its rivals describe as shooting itself in the foot—and consistently win over those same rivals.
Massive outages on AWS will force companies rethink their strategy.
China relationship causes massive rejection of made in china tech.
ARM's US listing will be aborted in favor of private sale to unreliable new owners, setting off an IP conflagration that no one can sort out, putting CHIP investments at risk and situating tech as ground zero for the economic & culture wars of the US 2024 election.
2) White separatists could take controll of a medium sized section of Northern Idaho and the standoff could turn into a stalemate.
3) Joe Biden could have a stoke yet still insist on running for a second term.
4) A sink hole caused by excessive pumping of ground water could cause a sink hole which at least temporarilly swallows a major river.