So I'm currently using "OpenCV University"'s playlist on YouTube to get myself up to speed with computer vision, and this has lead into a spiraling staircase down into the depths of CNNs.
And after that, I've had some recent projects that I love to mess around with such as a better license plate detection API than what currently exists for U.K. plates, and once I completed those two courses I had a good enough baseline to work from where I'd encounter a repository and google around if I needed to learn something new.
Short, simple, not painful etc. and I don't have the advanced mathematical background (nor the background within the American mathematical notation) that I'd need to digest the MIT course set, so this learning path has been the best for me. I'm no expert whatsoever, though.
If they can paint, then so can I. And it took me years to learn how to do decent stickmen!
Either way, we are not going to see intelligent computers in our lifetime, let alone elephants. Don't mean to condemn them but sapio-genesis is often oversold as being too easy. We are nowhere near being capable of ourselves, we have just learned how to process things similarly to how brains do it, at a fraction of the efficacy and ten trillion times the cost.
> we are not going to see intelligent computers in our lifetime
My dad was born before the first transistor was built, and died in 2014.
When he was born, reasonable people thought they'd never see a nuclear chain reaction, never see supersonic flight, never see space flight, never have men walking on the moon.
In his lifetime, even after the invention of the computer, reasonable people thought they would never in their lifetime have a machine beat the best human chess player. One of his specific anecdotes was about "fitting all of Shakespeare's works on a ball bearing".
Even as late as 2004, the script writers for "I, Robot" considered it sensible to have that memetic Will Smith line "Can a robot write a symphony? Can a robot turn a canvas into a beautiful masterpiece?" — my dad never saw Stable Diffusion or ChatGPT, but he would have if he'd lived the average life expectancy.
No matter what exactly it is that you mean by "intelligent" such that it excludes what computers already demonstrate, a lifetime is a long time, and a lot can change.
A lot of the velocity you listed was not as quick as it seems.
Supersonic flight is a question of manufacturing processes more than anything; we could always scale engines to make more power, just that making them light while being producible to a specific degree of accuracy was difficult.
Nuclear fission was discovered in 1938 and implemented successfully in 1945. Of course nobody would have thought of it - just like how I would never have thought of a ladder if I never saw one and never had a need for one. It's a question of looking at the fundamental laws and materials of matter, finding their inner workings and theorizing more useful ways of using these inner workings. We didn't put a stupid amount of effort into nuclear fission relative to how much it changed the world - I'm willing to bet that more cost has been put into r&d for office furniture than for the nuclear bomb.
Chess is a simple game of calculation, not a sign of intelligence. It merely reflects the ability to calculate, which is the strength of scalable computers. It just so happens that we came up with computers and they work well in this space - there's no telling of things we haven't discovered because we haven't thought to implement anything to do with them.
Sure, computers can write symphonies, but they are very far away from inventing something like a symphony. Which I would describe as something close to sapiogenesis.
I maintain that intelligence won't be cracked in my lifetime, knowing that I could look silly doing so. We can't even quantify what intelligence is, let alone begin to work towards it.
These things were all called "impossible" — and in some cases "unthinkable" — when my dad was born; supersonic simply looks like "a question of manufacturing processes" because we here in 2024 already know (as a society if not as individuals) how to make a jet engine that doesn't melt itself during operation and a rocket that doesn't explode. Going to the moon was itself the metaphor for "you can't do that thing". Rutherford famously dismissed the idea of extracting useful energy from nuclear reactions as talking "moonshine".
Before Deep Blue, sensible people insisted that Chess "required" the human intellect and couldn't be reduced to mere calculation. When it turned out to the contrary, then the goalposts moved to Go because that game was too complex to explore the state space and you can't even write a simple function to estimate how good a move was the way you can add up piece point values in chess. And then Go fell, and people suddenly decided it wasn't special after all, and "what really matters is emotional intelligence, which machines will never replicate":
If it is sufficient to look "at the fundamental laws and materials of matter, finding their inner workings and theorizing more useful ways of using these inner workings", then that would also suffice for intelligence, because we are existence proofs that it's possible to make a ~1.5kg, 20 watt, human-level chemical intelligence.
Everything we've done with A.I. since the field began, has been theory and testing of attempts to replicate various aspects of our intelligence.
> Sure, computers can write symphonies, but they are very far away from inventing something like a symphony. Which I would describe as something close to sapiogenesis.
If you shift the goalposts like that, then you would fail any single human, and only have sapience for the species as a whole super-organism. The word has meant a variety of different concepts before reaching its current meaning, as the economic and technological milieu, as well as the existing body of works that they were expected to be familiar with and innovate on (but only a bit because too much change all at once makes people hate you*) all continuously changed the composer's capacities for grandeur. No single composer created the modern form ex nihilo, nor even any single step towards it from the ancient Greek σύμφωνος (harmony), each step was building piece-by-piece on those who came before.
But again, this is to miss the point. The people in 2004 who wrote that script, didn't have Del Spooner say "Sure, we all know robots can write a symphony or turn a canvas into a beautiful masterpiece, but can they invent the concept of symphony if we never told it to them?"
To put it another way, using your own words:
> just like how I would never have thought of a ladder if I never saw one and never had a need for one
And yet, about 20 years later, never mind the 77 since the transistor or the 80.7 years of average UK life expectancy? The path the scriptwriters went for was to have the character of Sonny respond "Can you?", pointing out that most humans can't compose or paint either, not by making Sonny create works faster than most printers can actually print them — a thing which is already possible for high end systems today, which those script writers couldn't conceive of.
And not just script writers: between graduation (2006) and the actual release of these models, I've had conversations with co-workers, software engineers, who didn't think that what these models do was possible.
> I maintain that intelligence won't be cracked in my lifetime, knowing that I could look silly doing so. We can't even quantify what intelligence is, let alone begin to work towards it.
But we absolutely can quantify intelligence, that's what IQ tests are and we've had them for ages, the problem right now is that we're not sure if there's a single "general" thing or lots of small skills in different areas. Humans seem to have a G-factor, but the machines keep passing all our tests almost as fast as we can fill in the gaps in this sentence: "sure, they can ${old test}, but they won't be intelligent until they can ${new test}".
The Windows Internals series is typically the go-to for what you're looking for. Combine with MSDN browsing to determine what you want to do and how you want to do it.
But you have to be aware the part 1 is now a bit dated, the 7th edition is from 2017. The part 2 is from 2021. It’s not a big deal but some references are a bit outdated and some modern stuff is missing, but nothing that will be a blocker.
The books aren’t focused on programming though, but they explains the security, threading model, etc. Really interesting reads.
I comment it often but something great with windows API is that you can still use all the Win32 APIs, almost all old examples you can find on MSDN are still relevant to this day. The only thing you need to get started is a way to interact with C, and learn windows conventions (parameter names are weird at first but make sense, the error handling can also be confusing at first, the process model is a bit strange at first if you come from Unix, etc)
He does have a point, though. The Chinese labor policy of "more" is strictly benefiting the Chinese economy.
Not that it's much of a downside for us; if you live in the west you are competing with people who don't do much outside of 9-5. We have the benefit of having the ability to apply the Chinese "work till you're dead" mindset and gaining a competitive advantage, on an individual scale.
It all comes at great human cost of course, but it's not like it's not an effective method.
> Not that it's much of a downside for us; if you live in the west you are competing with people who don't do much outside of 9-5. We have the benefit of having the ability to apply the Chinese "work till you're dead" mindset and gaining a competitive advantage, on an individual scale.
The weird thing is that if you work at a company where this isn't the norm, then people will ask you to "slow down" or "take it easy". I believed them, until I realized that was their tactic to make sure you don't "steal" credit from them by working harder.
On the other hand, if you work at a company where "work always" is the mentality, then it's much harder to do that "one simple trick" where you get more work done by working more.
> until I realized that was their tactic to make sure you don't "steal" credit from them by working harder.
Definitely true from a office politics standpoint. And incredibly painful. Not all problems are solvable in the office and this is for sure one of them - when I was in this position, the only thing I could do was switch jobs.
Dunno, I got AWS CSA-P and it never got me any new interviews. Granted I stopped looking and only took interviews from people that looked for me actively.
> because the guy who did it out of hours was away at a picnic or whatever
I don't care if he was at home shitting in his hands and clapping. It's not his responsibility to update the user side of this product to the sole benefit of Amazon. That's where the story should end, IMO.
Started off here: https://www.youtube.com/watch?v=hZWgEPOVnuM&list=PL6e-Bu0cqf...
Ended up here: https://www.youtube.com/watch?v=_5XYLA2HLmo&list=PL6e-Bu0cqf...
And after that, I've had some recent projects that I love to mess around with such as a better license plate detection API than what currently exists for U.K. plates, and once I completed those two courses I had a good enough baseline to work from where I'd encounter a repository and google around if I needed to learn something new.
Short, simple, not painful etc. and I don't have the advanced mathematical background (nor the background within the American mathematical notation) that I'd need to digest the MIT course set, so this learning path has been the best for me. I'm no expert whatsoever, though.