Where is the AI we’re talking about?
When Clara invited me to give a talk on ChatGPT for artists, the first thing I did to prepare for this talk was to go to Dall.E 2, an AI that generates “art”.
After typing in “A Chinese Lady with a Pearl Earring by Johannes Vermeer”, I immediately got an image like this.
I like the result.
Then I tried “A lady dressed like ‘Court Ladies Wearing Flowered Headdresses’, with the ‘Dark Side of The Moon’ poster as background, by Zhou Fang”.
By the way, ‘Court Ladies Wearing Flowered Headdresses’ by Zhou Fang in 1200 years ago is one of the most famous Chinese paintings, equivalent to Da Vinci’s Monalisa.
I was disappointed because the result looks like the AI has no idea about my favourite band’s most famous album, ‘The Dark Side of the Moon’.
Of course, this is far from being ‘art’. The point is, with the help of AI, how quickly, how conveniently and how cheaply can a person with no art training create infinite images to present/express his idea?
This powerful AI is right in your pocket, and it’s getting more powerful every day.
I think the best way to understand ChatGPT is to think of ChatGPT as a renaissance man that has learned everything available on the Internet, from language to language, from text to music, from photo to movie, from science to art, from business to economics, from politics to philosophy, from psychology to theology, from plumping to biology, from sports to gaming, from gardening to space shuttle… including every word that people publish on the Internet and social media… And then this ChatGPT can talk to you about what it has learned, using any language you choose…
By the way, Google search gives you a menu, but ChatGPT gives you the answer. This is why Google’s stock price is falling after ChatGPT became popular.
Think about your call to your bank’s call centre, there is a non-AI system dealing with you. You can only choose from predefined options. It’s very much an ‘if/then/else’ theory. The programmer will pre-define what questions you can ask and what’s the answer to those questions. Anything out of that ‘pre-defined’ range is ignored, and then the conversation gets stuck and has to be transferred to my favourite option – ‘a human rep’.
But with ChatGPT there is no ‘pre-defined’, instead the ‘GPT’ stands for ‘Generative Pre-trained Transformer’. The creators of ChatGPT defined an artificial neural network, called Transformer, to deal with how to digest the input data, and use all the available data to train the Transformer, then the Transformer is ready to answer questions, and will keep using all the conversations as new input to keep training.
So your conversation with ChatGPT will have no range limits, and the conversation can go on and on, just like talking to a real person.
How do we get here?
When I studied computer science in college in the late 1980s, artificial intelligence was very different from what it is today. But artificial neural networks that copycat the human brain seem to be the most compelling future of AI.
In the epic chess games IBM DeepBlue defeated Kasparov in 1997, the technology was not neural networks. DeepBlue was still built on “if/then/else” theory, but equipped with purpose-built supercomputers. Kasparov was defeated not by AI, but by computing power. Therefore, people believed that DeepBlue’s approach couldn’t work in Go, the most complex strategy game.
In 2016, AlphaGo defeated Lee Sedol, the human world champion in Go. I feel his pain and human struggle when I watch the TV news. This time, the technology in AlphaGo is artificial neural networks, which is the similar technology used by ChatGPT and Dall.E 2 today. Now, ChatGPT is the first AI that has been trained by all possible content from the Internet. This scare of data makes ChatGPT the most powerful AI we have ever seen.
What’s the positive impacts?
Generally speaking, ChatGPT or any other form of AI is still a “tool” to empower people.
Office workers can use it to improve their job performance; musicians can use it to create new albums; visual artists can use it to create new works of art; scientists can use it to find new and better drugs; doctors and surgeries can use it to give patients new or better treatments …….
Skills may become easier to acquire as new training courses and devices are developed by AI.
Artists may be more creative as new ideas are sparked by AI.
Artwork may be more affordable to most people as AI improves the productivity of the entire art industry.
There are many entrepreneurs working hard around the clock to win the race of applying the new AI technology to their industry.
What’s the negative impacts?
What worries me about AI is that today the most powerful AI is already in your pocket, but we don’t really understand its power, or often underestimate it. Kind of like a toddler playing with a sharp knife. I feel like we are now in a period like when the car was invented, there are no road code, no traffic lights, no driving school, everyone can drive without taking a test. Then you give everybody a car. Yes, the key is in your pocket. Of course, this is crazy and chaotic.
People in the venture capital industry like me are usually optimistic about technology because that is how we make money. But when it comes to AI, I am really worried. Yes, I believe that we as humanity will eventually overcome this chaos period, but the problem is, how long will it take? How much will it cost? And who will pay? No one wants to be crushed by a new driver, but now the road is full of untrained and untested drivers.
The immediate impact, I and many people believe, is that the Internet and our world will soon be flooded with this AI-generated content. How do we recognize and deal with this new reality?
– Psychological Impact/Disorder? Our brains are not designed to deal with this kind of thing. People are more easily fooled and brainwashed by flooded AI-generated part true content.
– Job replacement/unemployment? All creative jobs including artists, writers, composers, and programmers are under pressure.
– Political instability/war? People will be divided by ideology, causing more conflict, even war.
– More unequal world? In human history, innovation and technology always make the world more unequal before it reaches a point where most people benefit. I hope this time, with AI, the world will become more equal or better as soon as possible.
Many people talk about how AI will enslave humans when it becomes ‘consciousness’. This is possible, but far from reality. I worry more about the evil of humans than the evil of AI in the future.
The problem that worries me the most is the impact of social media plus the power of AI. Since 2016, we’ve all learned how fragile our modern democracy is when it’s attacked by misuse of social media. How much worse would it be the next time someone uses the power of AI to launch more sophisticated social media attacks?
A weaponized AI targeting our democrat through social media is a real nightmare, no matter who is behind it.
Some questions worth thinking about with those flooded AI generated ‘arts’.
– Will this enhance an artist’s creativity, or weaken it?
– Will this make people love art more, or hate it more?
– Is an artist using AI a new race (AItist)? Or still a human artist?
– Who owns the copyright? AI, artist, or AItist?
– Will this encourage more people to learn art, or the opposite?
– Will this make artworks more valuable, or destroy the value?
What’s the take-away points?
- Be aware of the power of AI, talk to people about it, and be prepared for the coming impact.
- Explore the potential for using AI in your field for good, no matter how small the initial step.