This is the second article covering the report put out by a16z covering the state of the cryptocurrency industry. The first one can be found HERE.
In this we will go into AI. This, in my view, deserves a separate piece since there is a lot to what is taking place.
Image generated by Ideogram
State of Cryptocurrency: AI
The article, when digging into AI, references blockchain more than cryptocurrency. This is natural since we have seen this discussed on many occasions. In fact, we dive into it quite frequently. Personally, one of the keys to the future is the democratization of data. This is something that a blockchain like Hive can assist with.
At the top of the list with concerns regarding AI is centralization. We are really entering the world of Big Tech with chatbots and LLMs.
From the REPORT:
Given that the cost of training frontier AI models has increased at four-times per year over the past 10 years, we believe AI could trend toward increasing centralization of power on the Internet. Left unchecked, only the biggest tech companies may have the resources to train the latest AI models.
Could this be a situation where the "big get bigger". On the surface this seems likely.
As the training costs of models increases, that starts to limit the number of companies that can even consider embarking upon something like this. We are likely to the point where only a handful of global technology companies can muster up the resources to train a state of the art model.
Here is where blockchain enters. We have to democratize the process at every layer. As we discussed quite frequently, this starts with data. The focus upon processing is valid. However, without data, the compute that is acquired is worthless.
Of course, entities like OpenAi are enhanced by the fact that their existing models are prompted daily, generating more data. The question of how useful synthetic data is compared to human remains to be seen. Nevertheless, this is what we are up against.
The crypto industry, according to a16z, is already starting to move in this direction.
Crypto builders also have strong ties to AI. About a third of crypto projects — 34% — say they’re using AI, regardless of the category in which they’re building, up from 27% a year ago, per our builder energy dashboard. The most popular category for applied AI tech is blockchain infrastructure projects.
This is a very promising sign. As more platforms incorporate AI, this will result in those seeing an exponential rise in data. Also, over time, depending upon the design, we could see more available to other developers. Not only is the data permissionless (if written to a public blockchain) but we also have the algorithms and models being open in many instances.
Here we see Web 3.0 throwing its hat into the ring.
There are a few initiatives there were highlighted:
Crypto projects are already attempting to tackle some of these challenges today, including Gensyn (by democratizing access to AI compute), story (by tracking IP to help compensate creators), Near (by running AI on open source, user-owned protocols), and Starling Labs (by helping to verify the authenticity and provenance of digital media), to name a few.
Medium of exchange
Something that was not touched upon in the report and goes beyond blockchain is the idea of micropayments. It is here that I think AI and cryptocurrency are joined at the hip.
Read most forecasts regarding AI and it is evident that we are heading towards agency. The future of the Internet is tens of billions of AI agents. These are going to be handling many of the tasks that are now done by humans. Basically, if you can think of something that is done online, there will be an agent (many agents) that perform that activity.
As we know, a fair bit of what takes place is commercial This means payments are required. Agents are going to handle much of the shopping that takes place online, at least in terms of placing the orders. merchants, naturally, want to get paid.
An AI agent is not going to write a check, pay in cash, or swipe a Visa card. There could be certain applications tied to the likes of PayPal. That said, to me, crypto is the answer. The fact that crypto was developed in the digital world, unlike the banking system, means we have something that can scale with the pace of automation.
This is going to be very important.
The biggest factor leading to this conclusion is the fact we are often going to be dealing with micropayments.
Let us use the example of booking a hotel room.
An individual has a meeting in another town requiring an overnight stay. Once that is entered into the calendar, the personal AI agent goes to book the hotel room. This is not well trained on those types of sites (or more likely dealing with those agents). Hence, it interacts with a "hotel booking agent". The information is provided and the second agent handles it. For this service, the booking agent receives a small payment, say 25 or 50 cents.
Obviously, this is something that is not feasible with the existing financial system. To start, the operation of the system is too costly. In addition, intermediaries always want a cut. It gets very difficult to make these worthwhile when dealing with small amounts.
The hotel booking agent could be autonomously handing hundred of thousands of bookings per day. This is why it will have very low fees. It becomes a volume business.
Hence, for micropayments, crypto is the only real option. To take this concept one step further, consider the idea of an agent that only charges a couple cents for its service. How is Visa (or even PayPal) going to handle that?
We can see how in the AI agent world, cryptocurrency is required.
This is why an AI future must involved crypto.
They cannot be separated.