It's quite shocking to me how many experienced, tech-savvy people, who used to care about cookies and ad tracking - are now willingly sending their business strategies, highly confidential contracts, and intimate personal issues to a cloud provider because "it is only $0.0x per million tokens!".
> It's quite shocking to me how many experienced, tech-savvy people, who used to care about cookies and ad tracking - are now willingly sending their business strategies, highly confidential contracts, and intimate personal issues to a cloud provider because "it is only $0.0x per million tokens!".
Because there are more privacy guarantees there, depending on the provider. "But what if they violate their contract!" is some pretty tin-foil hat stuff.
How is this any different than a business running their website out of the cloud, assuming you are using a provider with appropriate contractual terms?
You can care about tracking and ads but still be comfortable storing your backups in the cloud, and many have been for quite awhile, even sometimes without encryption - that is totally different than e.g. Meta actively trying to track you and understand your relationship graph and your purchases etc.
I'm not sure about the tin-foil-hattedness of worrying about them violating their contract. But that's by-the-by. It is definitely not tin-foil-hat to worry about the data being taken in a breach.
Depends on how you use it. You can use AWS in a way that protects your data even if Amazon is compromised. You can even run LLM inference that way, but it doesn't seem at all common.
I would say tech workers have very weak class consciousness and understanding of power structures. Those who care about not ceeding power is a tiny minority compared to say among MDs and lawyers and their guilds/unions.
Ironically, the pill counting example selected to showcase "the best vision model" can be easily solved with OpenCV template matching, a technology created 25 years ago.
I'm assuming you mean that this tech became available in OpenCV 25 years ago, but as it turns out, the underlying tech can be traced back much further, at least as far as 1977! :)
https://ieeexplore.ieee.org/document/1674847
G. J. Vanderbrug and A. Rosenfeld, “Two-Stage Template Matching,” IEEE Transactions on Computers, Vol. C-26, No. 4, pp. 384–393, April 1977.
DOI: 10.1109/TC.1977.1674847
The point is that it's general. It can do this task and many other tasks and it doesn't need custom development like OpenCV does. Of course if you only want to count pills and you want it to be cheap/fast you're still better off using OpenCV.
Basic Template matching has severe limitations around scaling, rotation, and perspective. In my experience it greatly underperforms compared to deep network object detectors. My experience- and I imagine others have different experiences- is that SIFT techniques also fail pretty badly with noisy data.
That's correct, and I was specifically referring to the example chosen - where scale and perspective are known. Template rotation is relatively easy as well - but partial obstructions would pose a problem.
Another application where template matching would work brilliantly? Car counting in parking lots using satellite imagery.
Source: I did this [1] using OpenCV and template matching. Outperformed "Cars Overhead with Context" models.
I keep a card around like this. I use it once a year so they don't close it for inactivity, but it's a non-reward card so I don't use it day to day. It has a 8.50% APR from two decades ago, so I keep it around just in case. I can't really see where I'd actually need it, but it costs me nothing to keep it around.
There was a period during the late 2000s when credit card companies were offering 0% interest on balance transfers with no balance transfer fees, AND bonuses based on minimum card spend over a certain period after account opening. FatWallet (RIP) was full of people talking about how to use these offers to make loads of money in combination with rebates, eBay, bank account deposit bonus offers, and a lot more. Wild times. Long since over though.
Many years ago this "0% transfer" stuff was relatively new in the UK and a new outfit wanted to break into the market, they had two ideas which I guess they had costed as marketing ploys. First, the cards were a weird shape, this means there are a few applications where your card doesn't work, which is slightly annoying, but it's balanced by the brand recognition. Nobody cares which brand of rectangular plastic card you... oh, that's a weird shape.
But the Second was the easiest possible 0% transfer. To effect the transfer they write you a cheque for however much they'd agreed (let's say £1000) and you use that cheque to pay off a card or other line of credit. They charge 0% on this for 12 months.
What I, and lots of poor but money-savvy people did was sign up for the card. Deposit the cheque in an interet-bearing savings account, and set an alarm to pay the card off before that 0% expired.
So twelve months later you've made say £30 interest and you cut the card up. I don't know how many people did this, versus how many engaged with their product as they'd imagined. I know only two things:
1. I had about 50p outstanding balance to pay on my 12 month card, I figured they'd tell me I need to pay 50p within 30 days or whatever and if not they'd charge me extra - nope, they wrote saying "Your balance is negligible, we write off this tiny balance and don't expect to ever hear from you again".
2. This offer was never repeated. They did other 0% transfer offers but the "It's just a cheque" idea was never attempted again.
The frontier labs will do their best to prevent this from happening. Their financial model won't work if people start running open-weight models on their local hardware.
This is why banning Chinese open-weight AI models is a major policy debate in Washington. The labs can't survive log-term without subsidies, and a ban can act as a subsidy.
I concur, that this could be a possiblity and it definitely seems now.
Though my bet would be, if USA will go ultra protectionist in this regard - in 10-20 years most world will run Chinese LLMs and hardware for this purpose.
reply