Think about all the times in llm gets it wrong, the fact that would have helped to get it right is something that was lost. I suppose this isn't proof it's lossy just maybe we don't know how to get the data out.
Or look at it another way LLMs or just text prediction machines, whatever information doesn't help them predict the next token or conflicts with the likelihood of the next token is something that gets dropped.
Or look at it another way these things are often trained on the many terabytes of the internet yet even a 200 billion parameter network is 100 or 200 GB in size. So something is missing, and that is a way better compression ratio then the best known algorithms for lossless compression.
Or we can look at it another way, these things were never built to be lossless compression systems. We can know by looking at how these things are implemented that they don't retain everything they're trained on, they extract a bunch of statistics.
Just thinking that an unexpected event can't happen on the road shows a huge lapse in judgment and that's just the first sentence. Thinking they're special and different other people is another big red flag.
I was just using the price range provided when I google "ferrari price". The number range passed a sniff test so I picked the highest one and ran with it.
Your numbers and sources are clearly better.
Are there even 10,000 of those "hyper-cars" out there to be bought?
I am easy to classify as an anti-microsoft zealot and I am convince Gates is doing immeasurable good with his work in Africa against malaria. I haven't really looked into in the past year or so though.
Its almost as if they expect that just putting chips into a market makes that market have to buy it, almost as if they have gotten to used to not needing to compete. How could a chip vendor that has thoroughly cornered a few major market segments and might be considered to have some amount of monopoly status ever get into such situation?
So why do you say anti-competitive behavior wouldn't be unethical?
Sorry for the double negatives, I am just trying to understand. It seems clear to me that potentially hurting paying customers to increase one's own wealth is obviously unethical. This of course presumes malice, which hasn't yet been proven.
Saying "its a business decision" doesn't justify unethical and potentially illegal behavior. This is either incompetence or malice. If it is malice it falls well on the wrong side antitrust behavior. If its incompetence, well it isn't much better is it?
Doesn't giving web crawlers different pages than the user facing UI get you kicked out of search engine indexes? The original ticket indicated different pages were served, not just QOS.
I agree that other things might be possible, but looking a the history, assuming malice is a reasonable starting point with microsoft. It has been right many times in the past.
Sometimes companies with experience getting caught learn lessons on how not to be caught, rather than lessons on following the rules.
There is no reason with the modern setups that entirely different UI pages need to be sent based on user agent detection. Every best practice I have read says to detect for features, and fall back to something more general when a feature isn't present. Clearly that is not happening here, and that little bit of technical incompetence is enough room to slide in a few hundred mb/s of transfers for certain targets.
Microsoft has a long history of pretending to play and throwing curveballs. All the way back to Dr DOS. Their run in with the DOJ was just one time they got caught. They are even doing stuff now with CPU detection and not providing windows updates to people with and older but still "supported" OS and a newer CPU. Just because they won't let it go, they are still patent trolling android handset manufacturers.
Why do we think microsoft has changed for the better?
Or look at it another way LLMs or just text prediction machines, whatever information doesn't help them predict the next token or conflicts with the likelihood of the next token is something that gets dropped.
Or look at it another way these things are often trained on the many terabytes of the internet yet even a 200 billion parameter network is 100 or 200 GB in size. So something is missing, and that is a way better compression ratio then the best known algorithms for lossless compression.
Or we can look at it another way, these things were never built to be lossless compression systems. We can know by looking at how these things are implemented that they don't retain everything they're trained on, they extract a bunch of statistics.