Looking through the notes on new projects trying to “saddle” the blockchain, we often come across statements about the integration of AI into the technology of the blockchain. Let’s consider the issue in more detail and decide for ourselves how much such statements are real and what they have to themselves. Of course, this topic deserves more in-depth study. But within the framework of this article, we will walk through the visible tops.
- What is AI?
- Who needs whom?
What is AI?
There is no and there will never be a consensus on what to count AI. This is tantamount to determining what an intelligent person is. But there are some components of this ephemeral concept, concerning which you can somehow agree.
- The program should perform a certain function from the field of human activity no worse than a person.
- The program must perform a certain function from the field of human activity faster than a person.
- The program should be able to learn and improve its performance.
If the program fulfills these requirements, then we can consider this algorithm (program) to have at least some intellect.
From the whole conclusion. AI is just a concept and a guide. To date, there are programs, as components of something big in the boundless future, which partially solve the problems of automating the process of the human activity. They are the elements of the future complex and very intelligent AI. Once they are improved and combined into one big “organism”.
Examples? A lot of them. Starting from the calculator, ending with a robotic production line of complex aggregates. There are worthy results in the sphere of recognition of something. But it should be noted that they are achieved only by analyzing large data. Medicine, voice, video, pictures, search queries, mathematical operations. Separately, in each sphere, there are conditionally acceptable results, but before they are integrated into one they are still very far away. Even the seemingly acceptable results of chatting with bots do not solve any pressing problems. And here they are. My tests:
- Ask the machine — WHAT IS GOD? It will send you to Wikipedia or answer it with a standard phrase. Is this AI?
- Ask a question — A LEAN CAT COULD LEAN ON A SKUNK? All the programs that I asked this question were folded into a doll and instantly were forgotten by me. For there was no answer. The problem of homonymy and homogeneity is difficult to solve. And usually do not decide at all.
- WHAT WILL MIX 30% off GREEN and 40% RED and 25! % BLUE? Yes, I forgot to ask about 5% more. So let AI remind you.
All these tests are designed to show that before the creation of AI, oh how far. And the stuff that is presented in the press is just a hip for “hamsters”, over which professionals laugh.
Who needs whom?
Projects that go on sale of coins (tokens), usually are a continuation of some already working financial schemes and only. They are just trying to enter the crypto-currency market. To warm up interest in their activities, they declare the future availability of AI in their development. In this case, they have two options:
- Use your experience in this area.
- Borrow ready solutions from the “big brother”.
At first, the most difficult and attractive approach. It is expensive to have our own work. The project should have its own working model with demonstrated results. Developments in this field cost millions of dollars and are done for years. And if the company has achieved a certain result, then it understands how can use the product and where. I have not yet seen any mention of a project that, having its own developments in the field of AI, enters the market for blockchain technology (well, except that the project in which I myself am actively engaged. www.razoom.org)
Decisions from the “big brother”. Unambiguously working with them is easier, cheaper and more pleasant. But in no case can you build a reliable business model using other people’s solutions. Projects that use one of the dozens of free libraries for in-depth training are waiting for a lot of surprises. From the termination of product support by the developer, before increasing the cost of services. Plus, most importantly, the model used can very quickly become obsolete and lose in quality and speed to new developments. Well, do not forget that the “older brother” does not just give their developments for open use. Usually, such a maneuver is simply a desire to train the algorithm due to the free actions of users. Let us give some examples.
- Captcha. Very few people know, but when you unravel the captcha, we simply train the neural network to recognize the pictures.
- The mark of friends on photos in social networks operates on the same principle. I hope everyone remembers how badly this feature worked in the beginning, and how it works fine now.
So, using third-party services, we simply help them to improve and hardly control them enough to build on their platform any significant commercial model.
Blockchain. In short, this is a database that can be easily copied by everyone, but you can write to it using strictly certain tricky rules. So just described this technology is my friend and a very skilled programmer. Maybe it’s a surprise for you, but there’s nothing complicated in the technology of blockchain, and those mistakes that occur in projects with its use are just immaturity of the market.
To explain how it is possible to tie the existing components of the future AI to the blockchain, let’s draw an analogy. We have a torrent distribution (blockchain) and there is an interesting movie (transaction record), which everyone wants to get (put in a purse) with this distribution. Our AI should automatically analyze this film. He must find the names of the heroes, write a review using titles and a sound track, determine the category of the film, cover up the prohibited content and make a preview and a teaser. And of course, he should provide all this for those who use the film. Having made such an analogy, it becomes immediately clear that the flies are separate, but the cutlets are separate. We understand that doing an analysis of the content (film) is not the same as transferring it to everyone. But these two services need and can be merged, and get an interesting product. We need a separate service for sensing the meaning in the video, separately the service of video transmission and separately the service of recording the result of our work on the video. It’s like overlaying subtitles and voice acting. You can watch with her, and you can and in the original.
It becomes obvious that:
- None of the components (film, data overlay, data acquisition, film transfer) is not in the final product of the main.
- Only together do they acquire a unique value.
- It is not known which of the leading technologies.
- There is a distribution of the complexity of production and financial costs:
- Production of the film (content).
- Recognition of data in the film (AI work).
- Overlay the result (subtitles).
- Transfer of the finished product to the user.
1. In linking an AI element to a blockchain, the most difficult and costly process is the creation of an AI a component that will correctly solve the task.
2. The problem of large data. The film itself, which appears in our example (content), still need to get somewhere and do with it preliminary work:
- It should be interesting for the user to download to his “wallet”.
- It should have enough data for analysis by our AI component.
- The data must be correctly marked.
- Users should be enough.
3. Vectorization, categorization and markup of data. The technical question of the architecture of your AI component and the ability to perform complex calculations on a separate resource that is not related to the torrent itself (distribution). To date, this task is many times more complicated than the technology of the block.
4. So who needs whom? My answer is this. The blockchain is needed for the manufacturers of AI components, and not vice versa. He lets flies fly on cutlets and teach the producers of cutlets to make them “tastier”. All attempts to do the opposite, with a high probability, will not lead the project to a good result.
5. My advice. When you see the projects linking AI and blockchain, pay attention not to the blockchain, but to their AI component.
The result. Want to link the AI component to the blockchain?
- First produce or get somewhere “cutlet” (film).
- Do its marketing and distribution automation (let your AI do it).
- Lay it in the place where there are flies (decide your market fit).
- … and the flies will fly afterward (that’s where the blockchain is already).
If my story seemed interesting to you, then I promise to continue. Next time, let’s talk about NOT SMART CONTRACTS. Write in a personal, I will be happy to respond.