AI visionary: “Google has just gone bad in the meantime”

Richard Socher is fifteen minutes late for the zoom interview. He apologizes: an important meeting with investors took longer than planned. Socher needs powerful financiers because he has declared war on Big Tech. And that while the tech giants are in a crisis. Many are threatened with layoffs and their reputations tarnished. Small start-ups like Sochers You.com are now seeing their chance.

You founded You.com over a year ago. In that time, they have built a website that challenges both Google and the currently acclaimed chatbot ChatGPT. How can that be so fast?
Richard Socher: I sometimes wonder that too. We are only a small team, but with very good programmers. We all work very hard, but we are also pragmatic: we use existing technologies where available

You were born in Germany and have lived in Silicon Valley for years. Have you never thought about returning to Europe and founding You.com here?
I like life in the US. Not only because there are so many investors and programmers here, but also because the attitude of the people is completely different than in Europe. In Europe, I initially encountered a lot of skepticism about my ideas and people painted black what the AI ​​could do. In America, curiosity comes first and people say: Of course there are problems with AI, but let’s just solve them as best we can.​

They want to compete with Google with You.com. Are you megalomaniac?
Well, I think it’s a mix of weirdness and just the right timing to start something like this. Two years ago when me and Bryan[Bryan McCann, mede-oprichter van You.com, red.]started very small, most people thought there was nothing behind our idea. Interestingly, I recently read an interview with the founders of Google – and they were asked the exact same question

Where does your courage come from?
At this point, everything reminds me of my time as a PhD student when I worked on neural networks in language processing. At the time a lot of people asked me why I wasted my life on this, the concept didn’t work in the 90’s and won’t work in the future. Now all language processing is based on neural networks. I hope this is the second time in my life that I foresee such a trend

Google is one of the most powerful companies in the world. What else do you want to do?
Meanwhile, Google has just gone bad. The monopoly position and the focus on personalized advertising continue to hinder innovation. Google has a lot of interest in people clicking on the ad links at the top, because that’s how the company makes its money. You can’t just say, let’s replace these links with a chatbot that is much better than these links. So Google is very careful when implementing new features

Nevertheless, you can now get concrete answers to questions from Wikipedia and other sources. So you no longer have to click on a link.
40 to 60 percent of all Google searches are now so-called “zero clicks”. This means that you no longer click on a link to read the answer on another website, but the answer is presented directly on Google. You are not leaving Google anymore. This is useful for users. But it destroys the entire “internet ecosystem”. Therefore, this ecosystem must be open so that everyone can participate. At the same time, users must decide for themselves what they want to see, what not and how much data they want to disclose. You can do that on You.com.​

For example, if you ask ChatGPT about a person’s life, there will always be factual errors. A search engine that relies heavily on AI seems dangerous to us. How do you want to solve the problem?
Unlike ChatGPT, our bot always provides the source for such queries. This way you can check where the information comes from. That is very important to us. In addition, the AI ​​will continue to improve. Just a few weeks ago, people were telling me that there are no good answers and that everything AI does is nonsense. In the meantime, however, we have already solved a lot and it is our turn to make it even better

AI is not only smart, but also expensive. Just running ChatGPT would cost $100,000 a day. How do you want to cover the costs without placing personalized advertisements?
We are still working on our business model. At the moment we mainly focus on growth and reach, just like Facebook, Instagram and Youtube in the beginning. Some features are paid – such as generating texts such as emails from a certain amount. We’re not totally against advertising either, it just shouldn’t be based on the constant tracking of the user on the internet. Privacy is important to us.​

AI is also always criticized for reinforcing discrimination and bias based on the available training data. Do you have a patent recipe how this can be improved?
This is certainly a problem. Many AI researchers are now programming in their language models that the AI ​​will not even answer certain questions. I think that’s fatal! If you ask what the arguments are of the staunch opponents of abortion, they should not remain silent, but name the arguments. Also very extreme opinions. But it is also important that the AI ​​does not just reveal the most extreme opinion on its own

Let’s look to the future: how will AI change our lives ten years from now?
There will be a new wave of automation, as happened with the printing press, the steam engine and the Internet. The AI ​​helps to generate a large number of new ideas in a very short time. That changes advertising, communication, but also art. For example, a musician can generate several songs very quickly. But how good they are depends on his input and his intentions. People who understand how to use this new technology as a superpower will move forward. In the long run, I am very optimistic that AI will have a positive impact on our lives. But in the short term, the state needs to help people with social systems and offering further training

And education needs to change.
Teachers already realize that writing essays no longer makes much sense if students let ChatGPT write the text. So the way we learn and what we learn has to change. Of course, you still have to learn ABC, and you have to learn how to write without help. But you also have to learn to use the possibilities of AI. And of course you have to learn how to program.​

So why can the AI ​​also program?
If you can program, you can influence and further develop the AI

What else characterizes man in a world of intelligent machines?
It will always be the one that determines whether you like a book or not, for example. In addition, empathy is reserved for humans, although you can already have conversations with an AI today. You also always want recognition from people and not from a machine. All this will hardly change. And this is good news for psychiatrists

And finally, the question of all questions: Will machines ever evolve consciousness?
Oh, consciousness, that’s one of those things. First of all, everyone defines consciousness differently, from philosophers to religious scholars to anesthesiologists. But in the end, the question is not so interesting. Because in the sense that AI consciously sets itself goals that it wants to achieve, we are still a long way from that. And actually no one is doing it. Because what good is a machine that has to answer a question, but then says no, I’m going to do something else now. Such an AI would be completely useless and would only consume electricity. A sentient machine is ultimately a fun brain game for science fiction. Not anymore. (aargauerzeitung.ch).