You could actually add mazes and paths through them to the training corpus, or make a model for just solving mazes. I wonder how effective it would be, I’m sure someone has tried it. I doubt it would generalize enough to give the AI new visual reasoning capabilities beyond just solving mazes.
They could always satisfy that with an iPad or tablet? Also, I think it matters where your rear pillars are on if you need a backup camera or not, older cars have much bigger back windows but are more likely to kill you in a roll over.
“a small experiment on select Shorts, using traditional machine learning to clarify, reduce noise and improve overall video clarity—similar to what modern smartphones do when shooting video.”
It looks like quality cleanup, but I can’t imagine many creators aren’t using decent camera tech and editing software for shorts.
Well yes, that's what I mean, quality cleanup is not what I'd call a compression algorithm.
And as you say, arbitrarily applying quality cleanup is making assumptions of the quality and creative intent of the submitted videos. It would be one thing if creators were uploading raw camera frames to YouTube (which is what smartphone camera apps are receiving as input when shooting video), but applying that to videos that have already been edited/processed and vetted for release is stepping over a line to me. At the very least it should be opt-in (ideally with creators having the ability to preview the output before accepting to publish it).
I really have no idea how IBM is still in business, or the other big toxic techs like Oracle and Salesforce. Just goes to show I don’t know as much about the industry as I think.
Oracle basically runs HR and finance services for like every large company in Europe. They also run a scary amount of healthcare stuff and other government tech type stuff.
It sucks but I see why they do it. If you don't have the technical/managerial talent to handle procurement then it's the safest bet.
They bought Red Hat, which has OpenShift and all their other "DIY Cloud" bits. This stuff is popular in government or old businesses that may have been slow to (or unable to for regulatory reasons) jump to AWS/GCP etc.
To say nothing of the banks and others still using the IBM big iron.
The American hyper scalers are not necessarily the place to be. Modern can mean Non-hyper scalar as well. Can this sentiment just die please? Great that its working out for you and you replaced good sysadmins with aws admins, but it should not be the default strategy perse.
Why does this read like a personal attack? Do you have anything in my comment to refute?
I didn't even use the word "modern."
I actually agree the traditional cloud providers have lots of issues and aren't always the right choice, but the fact remains that offerings from Red Hat and the like are far more popular with older larger corporations than startups or "household name" tech companies like X, Netflix, etc.
IBM still sells mainframes and similar. And has a giant consulting and service business.
Their purchse of RedHat flows into consulting. Their purchase of Softlayer (rebranded into IBM Cloud) is more IBM owned, customer operated computing, a business IBM has been in since forever.
I felt like he avoided saying anything negative about Intel just in case it would be taken that way. Intel doesn’t have the best reputation so we are all interpolating a much more negative message than he actually said.
Agreed. He also mentioned these years being “some of the toughest at intel”. To me it read as 1) Amazing that he managed to get anything done at all with this kind of turmoil and 2) A not so subtle hint that things aren’t all good at Intel.
Good at self-promotion == just good in most cases for most practical purposes whether it's factual or not, arguably. His books seem substantial though, I don't know many people who've read or written 800 pages on system performance
> Good at self-promotion == just good in most cases for most practical purposes whether it's factual or not, arguably.
This does not seem true to me. Most popular programming YouTubers are demonstrably great at self-promotion but tend to be mediocre or bad programmers who know very little, even about the topics they talk about.
If anything we have plenty of examples of where being good at self-promotion correlates inversely with actual skill and knowledge.
With that said, I wouldn't classify Brendan Gregg as being good at self-promotion.
In terms of their compensation though, it functionally doesn't really matter, and that's somewhat true for being a professional as well, it's usually only important how many people think you're good enough. A job is often as or more political as it is technical
> This is a wild test, because LLMs get really pushy and insistent that the dog only has 4 legs.
Most human beings, if they see a dog that has 5 legs, will quickly think they are hallucinating and the dog really only has 4 legs, unless the fifth leg is really really obvious. It is weird how humans are biased like that:
1. You can look directly at something and not see it because your attention is focused elsewhere (on the expected four legs).
2. Our pre-existing knowledge (dogs have four legs) influences how we interpret visual information from the bottom-up.
3. Our brain actively filters out "unimportant" details that don't align with our expectations or the main "figure" of the dog.
Attention should fix this however, like if you ask the AI to count the number of legs the dog has specifically, it shouldn't go nuts.
A straight up "dumber" computer algorithm that isn't trained extensively on real and realistic image data is going to get this right more often than a transformer that was.
Yes, its all evolution. 5 legged dogs aren't very common, so we don't specifically look for them. Like we aren't looking for humans with six fingers.
I get it, the litmus test of parent is to show that the AI is smarter than a human, not as smart as a human. Can the AI recognize details that are difficult for normal people to see even though the AI has been trained on normal data like the humans have been.
I think the LLM is just trying to be useful, not omniscient. Binary thinkers are probably not going to be able to appreciate the difference, however.
If you want the AI to identify a dog, we are done. If you want the AI to identify subtle differences from reality, then you are going to have to use a different technique.
A lot of debt also arises because of savings needs. If everyone is saving for retirement, for example, that savings has to be debt marked somewhere else. Examples:
* Social security used to have a huge surplus, that was savings that had to go somewhere (even if it was just a savings account in a bank, the bank would then be able to lend it out). They instead buy treasuries and that savings becomes debt to the USG.
* China likewise needs to save dollars because it doesn't want them sloshing around in their economy leading to inflation, so instead of using it to buy things they buy treasuries, and their savings becomes debt to the USG (not always a great deal for China if interest rates are below inflation).
The dollar has been so useful in the past as a currency of trade because you could save large amounts of it easily by buying US treasuries. One reason China doesn't want the RMB to be used so heavily for trade is that they don't want to do the same yet.
reply