During the pandemic I made a list of predictions for the next year and when I revisited it I realised that I was expecting things to move faster than they did. I think some of them might come true eventually, but when is harder to determine.
I didn't expect AI to explode in the way it did; reading books like The Singularity Is Near and Superintelligence I was led to think that common AI tools would be self-learning and iterative systems that operate autonomously rather than prompting back and forth.
Here is the list I made in 2020:
• AI will become mature and we will interact with it as something with rights and individuality
• Stationary computers will be niche as laptops become powerful
• All newly built cars and some planes will be electric
• Phones will process everything in the cloud, the phone is a receiver/sender for information not a full computer, leading to cheaper/less powerful phones
• We will have clothes with digital aspects (colours, patterns, maybe textures)
• More (straight) boys will wear make up
• India will replace China as manufacturing powerhouse
• Asian culture will merge with Western, especially music
• Knowledge will be seen as more subjective and concept of privacy will disappear
While a formalized concept of 'turing completeness' which could prove the equality of sufficiently advanced LLMs will not exist yet, it will become clear that all models of a certain capability are equally capable(just as all CPUs are equally capable) and LLMs will become a comodity traded by speed, price and context window width. Furthermore, it will be realized, that beyond a certain size, larger context windows lead to diminishing output quality and we'll start talking about the theoretical limit for context window size (this will be a function over the amout of training data). OpenAI will still be a valulable company, but there will be at least 5 LLM providers to choose from and none will significantly outshine the rest.
At least one company will raise over $1billion on entirely fraudulent AI (using gpt's API and sleight of hand).
It doesn't need self awareness. Someone trained a bunch of AI girlfriends on her own personality. And Miku has been around for a while. And fake influencer technology. It's only a matter of time before someone does it.
Once all the information cannons for the US election start firing I think it will certainly seem anything but boring, but maybe not so much compared to a global pandemic.
The longer the AI bubble runs, the closer we get to defining where it's actually viable.
There are definitely some tricks it does well enough. I'm thinking of the old bit about the art students who make a better living than senior engineers on furry and fetish work, for example. That works because it's an industry where "good enough" pays the bills.
Conversely, industries where there's high standards, a risk of life/limb/finances, and associated regulations, are going to have a hard time with AI because it makes compliance and liability hard to manage. I could see some players trying to produce "HIPPA/PCI/etc-friendly hosted AI" or just putting up large insurance bonds to make their products amenable to industry, but there's a lot of "entire sector goes poof with the stroke of a regulator's pen" risk.
AI as "an augment" or as the primary user interface is at risk of trust corrosion. If the search engine/chatbot fails to respond reliably to prompts or hallucinates one wrong fact too many, the customer is going to bail. This has been a problem since Siri, if not every annoying IVR service-- we have marketers who desperately want to present the machine as open-ended and omniscient, which leads to insufficient documentation, poorly bounded user experience, and no easy way to control the user's expectations.
Many technical advancements are about being able to replace a skilled artisan with a less-trained and cheaper worker. I think a lot of the "AI for writing and coding" stories are going to fall flat there. You didn't get rid of the junior coder, you turned him into full-time QA, code reviewer and project manager to try to steer the robot, which is a lot harder!
I expect by the end of 2024, we'll have seen a fair number of services scale back the AI features, or at least constrain them-- offering AI interfaces alongside conventional ones when you absolutely need control. I don't doubt there will be some players with infinite budget still pushing it (Microsoft trying to make AI a thing may well be much like Meta trying to make VR a thing) but I don't think we'll be seeing so many 'your morning danish... NOW WITH AI' pile-ons.
yes, any reasonable person would bet on AI stocks in companies that have a real stake in AI trends that will reasonably lead to future productivity gains and aren't just using it for clout or PR
- AI will enter another ice age, rather unexpectedly, after it publicly contributes to something particularly egregious. Maybe it’s discovered that {company} helped generate a bunch of fake political ads (including faked audio and video)
- The US will withdraw significant financial and military support from Ukraine
- … in order to support Taiwan during a conflict that will break out in Spring 2024
- Significant FDI will migrate from China to India and Southeast Asia during ensuing conflict. China will experience a recession and social upheaval by Summer 2024 as a result.
Not long ago, Xi said, "Right now there are changes, the likes of which we haven't seen for 100 years..."
I think he's referring to the establishment of BRICS and neutral settlement, and he sees the days of the dollar(petrodollar) as waning. But there are huge ramifications to that.
Wouldn't that make his predictions even more significant? He controls the biggest population, second biggest economy in the world, one of the fastest growing economies and middle classes out there, has one of world's biggest network of spies feeding him info internally and externally. If anyone could manifest his fantasies, it would be Xi.
(I predict csomar will create another one in a couple of weeks ;))
I also predict most of the low effort, ChatGPT-based businesses will shutter, and companies will sunset ChatGPT-based features because they're not worth maintaining. There will be no major disruption/breakthrough in the technology, the hype will continue to cool as the average person moves on to the next thing. The market will continue to suck, but not because AI is taking our jobs.
Barry-1[1] manages to climb 100 Km higher in orbit without using propellant. Quantized Inertia, on which the purely electrical thruster is based, starts to draw interest.
You can follow along, watching the Semi-Major axis. There are other experiments on board, once those are complete, they'll be switching on in a month or so.
When I recently was in Thailand and used Bolt to get taxis, I could write in Swedish and they in Thai and the app would translate to our native languages in my experience perfectly to arrange pickup location and logistical details.
I think it is entirely realistic that some big or prominent companies will start to go away if there isn't large crisis which leads to drastic dropping of the rates...
a major class action lawsuit about the use of leetcode style interviews as a closet tool for age, gender and race discrimination will be won against a big tech co
i guess you can argue for age discrimination but come on, what does leetcode style interviews have to do with race and gender discrimination? In some sense, these interviews are exactly the type of interviews you want to not have race / gender bias since they are objective (either you solved it or you didn't).
Because leetcode-style interviews don't measure if you're good at your job; they measure the absolute entry bar for developers: the fact that they're comfortable enough to solve simple, abstract problems using a programming language of their choice. I'm not sure why would you want to remove that entry bar legally, but working in a team of developers that are not proficient with code sounds like a bad idea. Unless you envision a team in which a couple of people do all the coding and the others just have "good ideas", which is an absolute, utter nightmare.
I'm curious why leetcode-style interviews should be a proxy of gender / age / race though. Only old white guys are allowed to leetcode?
not all problems in leetcode style interviews are "simple" and most of them cannot be solved in a few minutes (which is the time you have).
these interviews are being used by employers to discriminate people however they want. they can just say you didn't do the coding interview well enough while what they are really thinking is, don't want someone from europe, too old, don't want a female, don't want a black guy etc.
also, people go to university and get a CS degree exactly to not have to go through this bullshit. and in the US, people pay off hundreds of thousands of dollars in student debt for their degrees. people who manage to graduate in a good CS program can pick up any language or toolkit in a week or two max.
lastly there is no correlation between the work you actually end up doing on the job and the coding interviews.
> lastly there is no correlation between the work you actually end up doing on the job and the coding interviews.
This is true but irrelevant: leetcode tests are just a first level cutoff to separate people who can code from people who can't. If you can't code you surely can't do the job. If you can, you might be able to do the job.
> they can just say you didn't do the coding interview well enough while what they are really thinking is
They can already say that you're not a good cultural fit or that they found someone else. At least leetcode-style testing is objectively measurable in some respect (the tests work at the end or they don't) and it's easy to record (some platforms record sessions by default)
> also, people go to university and get a CS degree exactly to not have to go through this bullshit
What do you mean? That people get CS degrees to never have to solve abstract coding problems? If anything, they get CS degrees to learn how to solve abstract coding problems efficiently.
> people who manage to graduate in a good CS program can pick up any language or toolkit in a week or two max.
Then no problem - they can pick up leetcode problems in a couple of days!
> - Ukraine will be forced to start negotiations on giving up land to Russia.
Unfortunately, you probably right. Very bright seen in Ukraine, Korean scenario.
Most, West could do at the moment, to convince Russia, to not include Ukrainian territories into their official papers, but territories will remain under occupation.
I'm most worried about rising sea levels. Climate change was a major election item here because we're getting massive floods every year.
Populist parties here promise world changing solutions - clean energy, carbon neutrality, etc. The more pragmatic ones suggest flood mitigation, mangrove forest coverage, still more sustainable energy but less aggressively.
Well people, you have really interest predictions, but they all depend on how Ukraine withstand.
This is not very high probability case, but unfortunately, it is really possible, Ukraine could fall.
If Ukraine will fall, will start chain reaction of conflicts in all parts of world, and US will become literally owe their allies to help.
And this is not only about tiny countries, this will involve China-Taiwan, China-Korea, and may be even China-Japan, and Putin will not stop in Ukraine but will continue in Baltic and probably, Poland.
If will happen so many wars, in 2024, probably, US will consider to turn economy to war rails, and next years could become similar to 1940th, when military administration dictated to business, what to produce.
For example in 1940s, already exists good diesel locomotives, but mils prohibited steam locomotives producers to switch to diesel, and huge part of diesel production directed to ship building.
Unfortunately, if Ukraine will just stop on current positions, or even if Ukraine with western help will return territory from occupation, this will not mean world is safe.
- Russia constantly grow military production, and even when we see limits, we cannot be sure,they will not invent some new method to avoid sanctions. And even with current production, they could attack EU in 5-6 years, so now, in NATO conduct talks about be ready for war in 3 years.
So, any way, in 24, Ukraine war will constant headache of west, and US military already officially said, they preparing for war with China in 25, and I'm to lazy to list how Poland and Baltic countries already preparing for war, you could read in wiki and google.
- I predict Trump will say the quiet part out loud: these elections are annoying and won’t you be relieved that 2024 is the last one?
- The plan to purge the Executive branch is well-defined for Heritage Foundation causes, too boring and preachy for Trump. He enacts some of it, but mostly offends true conservatives by irrationally attacking things like Federal accounting standards and postal carrier pensions.
- The biggest disagreement between the Heritage Foundation and Trump’s followers will be the single statute they fear most: the Hatch Act. Trump himself won’t care, and the battle to neuter the Hatch Act will elevate an insider to be his successor, because:
- Among his loyalists, the spoils system: identify a civil servant whose job you want, hound them and demand in social media your immediate appointment. I predict Trump delegates this not to underlings but to social media itself.
- The value of spoils here is the delusional interpretation of Federal statutes criminalizing interference of Federal employees. It is not a bold prediction to say these people are immune to shame. An IRS appointee just walks onto flights for free. A Department of Agriculture appointee takes his cameraphone into the women’s locker room. An immigration and naturalization appointee runs a beauty contest of immigrants. The point of Republicans tolerating this will be that it’s the fastest, dramatic solution to prove we need to shrink the evil government.
I didn't expect AI to explode in the way it did; reading books like The Singularity Is Near and Superintelligence I was led to think that common AI tools would be self-learning and iterative systems that operate autonomously rather than prompting back and forth.
Here is the list I made in 2020:
• AI will become mature and we will interact with it as something with rights and individuality
• Stationary computers will be niche as laptops become powerful
• Music production on phones common
• Phone connection points disappear (headphones, SIM card, SD card)
• All newly built cars and some planes will be electric
• Phones will process everything in the cloud, the phone is a receiver/sender for information not a full computer, leading to cheaper/less powerful phones
• We will have clothes with digital aspects (colours, patterns, maybe textures)
• More (straight) boys will wear make up
• India will replace China as manufacturing powerhouse
• Asian culture will merge with Western, especially music
• Knowledge will be seen as more subjective and concept of privacy will disappear