DeepMind's new Gemini AI will combine LLMs with techniques from AlphaGo

That’s wrong though, no matter how many people on HN repeat this.

First of all, an open protocol like that can enable global payments and other interactions without having to deal with tons of individual banking APIs or some trusted intermediary like Stripe. Like when the Web (HTTP open protocol) allowed people to publish a website in one part of the world and have people in another part of the world get it very shortly afterwards. It allowed commoditizing any middleman into a dumb hub (eg VOIP brought the cost of phonecalls down to near zero).

When any community grows large enough that its digital assets and representations have significant value (the outcome of elections, some roles in the community, its currency, access to some real-world goods etc) then a blockchain becomes important. Otherwise we have to again trust a single entity, and the way many US states deal with this is requiring this entity to register in every state and provide a surety bond, to make sure it really pays out. This applies to Money Transmitters (FINCEN) and also applies to Sweepstakes (which are almost like lotteries but have no purchase necessary) and other things like that. On a blockchain, none of that is necessary, nor are you forced to trust centralized entities. We see what happened with FTX and Celsius and other centralized entities, that were the complete opposite of blockchain but tainted the reputation of blockchain. People conflate it out of ignorance — but HN denizens know better. They may do it out of different motivations.

Finally, there are many places in the world where people do not trust their banking systems and governments to take care of their money. The banks rehypothicate and the governments print it, but in parts of the world, people lose much of their money. It is good to have an alternative. Places where payment and remittance systems are not very robust, have been seeing quite a lot of blockchain use.

Then there are logistics and other areas where multiple mutually distrusting parties want to secure a database together, and have a single source of truth.

It is simply not true that blockchains have no valuable applications. And here is the thing… the downside of all these applications is capped, and in the absolute worst case is limited to the value people voluntarily decide to commit to them.

By contrast, the downside of AI can be human extinction — with a probability above 10% if their own top people are to be believed. And humans have tended to underestimate what AI can do.

Even lesser scenarios like simply destroying trust in anything published on the Internet (ie all information, basically), or merely destroying the livelihoods of artists, writers, coders and others affects people who never voluntarily chose to put their value into the system.

So even if Blockchain had very few valuable upsides (which I argue is not true), its downsides are far smaller than AI’s downsides. It’s irresponsible and reckless to develop better and better AI and give it to everyone — but the same is not true of blockchains!