Chrome already has optional built-in support for generating alt text for images. It's been there for years, using a server-based API.
It does seem possible that this could be replaced with a local model in the near future. It's not clear the average user has the hardware specs for this to be an option today, but it will increasingly be plausible.
Keep in mind, though, that alt text is just one small part of making a web site accessible.
> It does seem possible that this could be replaced with a local model in the near future. It's not clear the average user has the hardware specs for this to be an option today, but it will increasingly be plausible.
Siri does something like this when reading messages into your AirPods. It will give brief descriptions of photos sent in the message. I'm pretty sure it's all run locally.
I think they're saying is that in Google's very early days, they gained market dominance against a dozen other search engines simply by being better. It wasn't until years later that they started paying to be the dominant search engine.
> It also impedes accessibility-related plugins trying to extract the content and present it to the user in whatever way is compatible to their needs.
I'm not sure I agree that this is relevant advice today. Screen readers and other assistive technology fully support dynamic content in web pages, and have for years.
Yes, it's good for sites to provide content without JavaScript where possible. But don't make the mistake of conflating the "without JavaScript" version with the accessible version.
Readers for the blind not the only form of assistive technologies, and unnecessary JS usage where JS is not necessary makes it hard to develop new ones.
There is a huge spectrum of needs in-between, that LLMs will help fulfill. For example it can be even as simple as needing paraphrasing of each section at the top, removing triggering textual content, translating fancy English to simple English, answering voice questions about the text like "how many tablespoons of olive oil", etc.
These are all assistive technologies that would highly benefit from having static text be static.
I think it'd also be worth comparing it to cars that have more basic collision-avoidance.
I don't expect Tesla's FSD to be perfect. But I do expect it to not be worse than the simple and effective collision avoidance that nearly all new cars have.
I sincerely hope this is just marketing, because none of the actual properties of blockchain make any sense whatsoever for this type of application.
Blockchain is decentralized. Why would you want that property for vehicle ownership that's specific to one department of a particular U.S. state? Having servers around the world sounds like a security risk, not an actual benefit to Californians.
Blockchain transactions are permanent. How is that helpful if a title is stolen, or transferred fraudulently, or illegally? What if a transaction contains sensitive or private information, or inappropriate language?
Without decentralization, doesn't that mean you don't even need a 51% attack... the central authority can just rewrite the ledger history at will, the same as any database or git history?
Just a different way of having an auditable and verifiable transaction history. Maybe it’s easier to share the blockchain than it would be to share copies of the database? They could be leveraging the existing tools for copying and verifying it, perhaps one of the requirements is that end users or dealers need to have access to a local copy that they can use to check the ownership history of a vehicle before a purchase? Everyone loves to hate on anything blockchain but it’s a useful data structure.
I don't think this really makes sense...? A blockchain is awesome for decentralized verification of a shared ledger, but without that, it's no different than a line by line history of writes in a text file.
If a dealer has a local copy of the ledger but isn't really part of the blockchain consensus, it's no different than having a read only history in any other structure.
It doesn't matter what format you store it in, it's more a question of who can rewrite history. If the central authority can do so at will and nobody can disagree with them, it nullifies the benefits of using a blockchain.
The dealer can say "it doesn't match our records", sure, but they can say that no matter if a blockchain or a bunch of CSVs are used. If the DMV then says sorry, our records are the only authoritative ones, please overwrite yours... then we're right back where we started.
I don't think people hate blockchains as a data structure, but as misapplied snake oil that doesn't solve the underlying problems. This doesn't create a real audit trail as much as the illusion of one. A proper audit trail would require the authoritative history to reside elsewhere, in some independent third party that the DMV cannot unilaterally overwrite.
They’re functionally equivalent these days. Blockchain has become marketing speak mostly.
But if you squint a little. The history of a git branch sure looks a lot like a blockchain (or vice versa). You have a ledger of records (commits), that are append only, and each commit contains a merkle or hash tree (git tree object)
I have to squint pretty hard to convince myself. Git itself really only covers the auditing part, which any static service will do by virtue of storing data. Git repos aren't decentralized, they rely entirely on external software like GPG and SSH to enforce identity and have no native consensus mechanism. They also don't protect from erasing transactional history which is kinda seen as the cardinal sin of ledger-based economics.
If you wanted to be reductive you could try to argue that all linked lists or hashmaps are a blockchain once compiled. I think that's silly, bordering on revisionism. Git wasn't programmed as a Blockchain, and Blockchains weren't intended to resemble Git.
Sure they are. Git is one of the pioneering “distributed” version control systems. Every clone of a git repo is a complete copy. Sounds pretty decentralized to me.
Git’s native consensus mechanism is to defer to humans and record their decisions as merge commits.
I agree git natively doesn’t protect against erasing history. Although rebase is discouraged and any one node can’t forcibly erase history from every other node.
> Every clone of a git repo is a complete copy. Sounds pretty decentralized to me.
Every clone of a DNS lookup table is a complete copy too. That doesn't change the fact that DNS is one of the least decentralized technologies being developed and propagated today.
It wasn't a file full of zeros that caused the problem.
While some affected users did have a file full of zeros, that was actually a result of the system in the process of trying to download an update, and not the version of the file that caused the crash.
Presumably you could roll out to 1% and report issues back to the vendor before the update was applied to the last 99%. So a headache but not "stop the world and reboot" levels of hassle.
Those eager would take it immediately, those conservative would wait (and be celebrated by C-suite later when SHTF). Still a much better scenario than what happened.
That's the thing, it's not very expensive! It's actually a pretty excellent price for what it does well. That's why so many people bought them and continue to use them.
Canvas would still be faster even if you used a full-screen box. Just the string concatenation overhead of doing this with box-shadows is insanely wasteful.
Which isn't to demerit the hackish creativity of taking one thing and running with it! But if you wanted to do a ball painting effect like that outside this "what if" context, it would be technically irresponsible to do it with box-shadows.
It does seem possible that this could be replaced with a local model in the near future. It's not clear the average user has the hardware specs for this to be an option today, but it will increasingly be plausible.
Keep in mind, though, that alt text is just one small part of making a web site accessible.