Thoughts on AI
Mapping current popular advances in AI (GPT3 et al.) to the present and the future.
Subject Area: Technology
We are Little Giants – Long Live the Capitalist? – Hollywood fooled us – The progress that scares – Loss of Value - A Dostoevskian Take
The recent massive advances of AI in the manner of chatGPT, GPT-3, et cetera (although, scarily enough, regarded as not particularly innovative by a brilliant insider) have brought to the fore, if anything, the future of humanity. I have registered some of my thoughts with friends, and I figured writing them down would be a good exercise.
We are Little Giants
Needless to say, the gap between our moral strength and technological prowess is getting larger by the day. And one is tempted to say that the “blessing” we have for now is that our superhuman (tech) advances are still somewhat restricted to the world of bits. But that would be cruel to say, given the litany of problems we will be able to solve with the bio revolution. The big revolution coming (it has started, actually) in synthetic biology would not be business as usual.
Synthetic biology is about digitizing and engineering life, advanced genetic engineering, if you will. And to think of how it might evolve, consider the current digital revolution, of which the AI revolution is a part.
Starting with computer access to only elite groups such as the military and universities; to the invention of the internet, the rise of home computers, Web 1.0, Web 2.0, social media, smartphones, and now the current advances in AI, Web 3.0, and blockchain technology.
Synthetic biology is starting to toll that lane. And biology is embedded so much more in our lives than we would initially intuitively think. The food we eat, the medicines we take, our pets, and even ourselves. It’s all biology.
There are so many examples to cite, and since I don’t intend this to be a biology essay, I will restrict myself to two recent exciting studies I came across 1) a new tool, PASTE, that can safely and efficiently snip out faulty genes and replace them with new ones, with potential applications in treating diseases caused by large numbers of mutations. 2) David Sinclair’s work on reversing aging in mice.
But we shouldn’t be naïve, should we? What about the potential risk of deleterious environmental impact and the accidental or even intentional release of bioweapons, to name a few. I don’t mean to be a harbinger of bad news; nevertheless, this is an excellent junction to cite Murphy’s law: “anything that can go wrong will go wrong.”
All this talk also reminds me of a quote from CS Lewis’s Abolition of Man:
“…There neither is nor can be any simple increase of power on man’s side. Each new power won by man is a power over man as well. Each advance leaves him weaker as well as stronger.”
Back to AI, here is a ChatGPT receipt: cybercriminals are starting to use it.
Long Live the Capitalist?
One of the arguments against the capitalists of our day would be (is being) proved demonstrably wrong in a roundabout fashion. Namely, the CEO/founder should not be rewarded as much as they are because they are riding off the back of cheap labor.
With robotics and AI/ML innovations, it will become conceivable for innovators to create billion-dollar companies with barely any human labor. A case in point is OpenAI itself, the company has about ~375 employees, and the value of that company is about $29 billion.
There are already a lot of indie hackers who make millions over the internet with practically no full-time employees. The question is, when will we get a $1B 1-person startup? More power will shift from labor to capital. Class struggle will intensify.
Hollywood fooled us.
The picture painted in the mind of an average joe is that blue-collar jobs will be obliterated first (and quickly) by AI or the sterotype of evil robots by Hollywood. But look at what we have. Large language models like chatGPT and its variants are coming for white-collar jobs and fast. But to be truthful, it isn't the case that Hollywood fooled us. Perhaps it’s just very hard to predict the future.
It should be noted that no technology in modern history has led to mass job loss among educated workers. What we have before us today is a technology with such potential.
The progress that scares.
When I was trying to pitch to my Ph.D. committee members why I wanted to switch my thesis from core biochemistry to a bit of machine learning, one of the arguments I used back then (which I borrowed from Andrew Ng) was that AI is the new electricity.
After a few years, I quickly realized that the analogy fails in at least a critical way. For example, the rate of adoption of electricity and the various innovations that electricity drives pales compared to what we are seeing with AI. It is not the progress in AI that scares me sometimes. It is the utter speed at which these things happen, and the scale at which it gets deployed. All you need is to just get plugged into the internet. Case in point: OpenAI acquired 1 million customers only after five days of pushing out chatGPT. That’s got to be a record or something. (update: In two months that number became 100 million active customers, with chatGPT becoming the fastest consumer app in history.)
Today, artists who have trained their entire lives to be good at what they do wake up ‘one morning’ only to be screwed over by text-to-image AI systems like Dalle and mid-journey.
I tried out some of these systems about four months ago, they are pretty good. See here.
And more.
So it’s little wonder that some artists have dragged some of these AI companies to court (this is because the AI was trained to ‘draw’ using tonnes of images, some of which belong to those artists). Interesting to see how that will turn out.
It is obvious that the revolution isn’t going to end with AI-generated text or images. Just a few days ago, Google Research published a paper introducing a model called MusicLM that allows you to generate music from just text. Yes, you heard that right. Check it out here.
Loss of Value
Growing up, the sight of a brand-new car was often memorable because most cars on the street were second-hand (we call them Tokunbo cars in Southern Nigeria). But the moment I came to the United States, virtually all brand-new cars ‘lost their value.’ Why? Because it’s ubiquitous and it’s becoming less memorable.
And what do new cars have to do with AI? The moment we can create text and artworks with AI, many human outputs in these categories will lose their perceptive value.
Low-grade writing powered by AI will flood blogs with images of people and things that never existed and will never exist (I am sure that is already happening.) I wager that this is a good time to invest in more physical things.
Talking about the loss of (perceptive) value, it is not happenstance that iPhone releases a new phone model every year. Let’s face it, who in the West or a developed country is in awe of her iPhone? No one. I am sure some folks somewhere are already getting bored of chatGPT
This leads me to my final point: aren’t we all moving in circles?
A Dostoevskian Take
Here is Dostoevsky in Notes from the Underground. I quote.
“Granted that man does nothing but seek that mathematical certainty, he traverses oceans, sacrifices his life in the quest, but to succeed, really to find it, dreads, I assure you. He feels that when he has found it there will be nothing for him to look for. When workmen have finished their work they do at least receive their pay, they go to the tavern, then they are taken to the police-station--and there is occupation for a week. But where can man go? Anyway, one can observe a certain awkwardness about him when he has attained such objects. He loves the process of attaining, but does not quite like to have attained, and that, of course, is very absurd.”
Close quote.
There are more worth quoting at some length, as these insights are piercingly intense, but let’s do just one more and call it a day.
“Then--this is all what you say--new economic relations will be established, all ready-made and worked out with mathematical exactitude, so that every possible question will vanish in the twinkling of an eye, simply because every possible answer to it will be provided. Then the "Palace of Crystal" will be built. Then ... In fact, those will be halcyon days. Of course there is no guaranteeing (this is my comment) that it will not be, for instance, frightfully dull then (for what will one have to do when everything will be calculated and tabulated), but on the other hand everything will be extraordinarily rational. Of course boredom may lead you to anything. It is boredom sets one sticking golden pins into people, but all that would not matter. What is bad (this is my comment again) is that I dare say people will be thankful for the gold pins then.”
This observation has been further articulated by modern psychologists in the manner of the hedonic treadmill, focusing illusion et al. (for more on this, see my essay “how (not) to be happy”).
Even if we are moving in circles, at least we take solace in that productivity (economic growth) is a powerful antidote to violence and all sorts of deformity - but not when such productivity exclude humans.
[Some notes:
[1]: MBB microgrant batch 2 is closed, and the grant winners will be announced towards the end of February. If you want to be informed when the next batch is live, you can place your email here.
[2]: I built an app over the winter break and would like to know if it’s worth putting more effort into. It works best on a laptop browser for now. Currently, you can autogenerate meal plans, grocery lists, and recipes (mealplanner.streamlit.app). Kindly give me any feedback you might have (there is a button to do just that in the sidebar.) Particularly, answer the question: for a subscription fee of, say, $1 or $5 every month, what new features would you like to see]
Thanks for sharing. Reading this has made me to go back to listen to John Gray's Talk titled the 'Abolition of Man' mirroring Lewis' book.
The problem I have with all the progress in AI is that no one can predict where it's going. Inventors and scientist can be working hard in the lab and computers on how these things will change our lives for the better, but they can't also police how others who would want to do us harm use it. T