Existential Risk from Super intelligent Machines

https://app.mindstone.com/mindspace/community/annotate/article_Agftqkx3Xmt7Snxioz?back=/mindspace/community/events/event/mindstone_toronto_april_ai_meetup



Sometime in mid 2023 Sir Geoffry Hilton announced his resignation from Google to be able to speak freely about existential dangers from AI. For those of us who may not know, Geoffry Hilton is the fatherly figure for Neural Networks. After the first AI winter, when enthusiasm and investment in AI had faded, He continued his research on neural networks, with support from his colleagues at the University of Toronto. In 2018 he and Yoshoa Bengio received the Turing Award for their work. The Turing Award is the equivalent of the Nobel prize for a computer scientist. He also said, “It is not inconceivable that AI could wipe out humanity”. He started a petition to stop AI development and In mid-2023 a group of very he along with other prominent individuals like Bill Gates and Sam Altman signed this statement of AI risk. All of this suggests that we should take their fears seriously. Dr. Andrew NG called for open discussions and I found most arguments favoured anthropomorphized projections of AI. So as a data scientist, I started enquiring about the survival of species and I looked at biology and the history of evolution for answers. Let me start with a story. 


A long time ago there lived two species, let’s call them species A and species B. Species A supremely intelligent and species B was super dumb. Compared to species A, B was 1000 times dumber. They lived happily as long as their ecosystems did not interfere. Around a few hundred years ago as human activity increased, their habitats collided. And soon there came a point where species A could no longer co-exist with species B. So, Species A actively starts to hunt down and exterminate Species B. How many Generations would it have taken for Species B to extinct? Species B is Blatella Germanica. Species A is Homo-sapiens. Us, Humans






Blattella Germanica is nothing but a cockroach. We have been trying to kill and exterminate cockroaches for almost 200 years now. Cockroaches have a brain size of 0.0005 grams compared to our brain which is 1200 grams. They have 100000 neurons we have 86 billion neuron connections. Just for you to get the scale. 100000 meters is 100 km that is from here to Niagara. And 86 billion is from here to the Moon and back 100 times over. So we are far far more intelligent than cockroaches and we are currently spending half a billion$ a year just to exterminate cockroaches. We have used Chemicals, pesticides, and biological control, we have even tried laser guns to eliminate cockroaches. And yet, after 350 generations cockroaches thrive amid humans. The fact that the cockroach control industry is expected to grow at 5% CAGR post-inflation is a testament to their resilience. There are over 4000 species of cockroaches only about 6 of these are considered pests and they are found only where humans live. For example, Blattella Germanica cannot be found in the wild living independently of humans. It is the same case with mice. We have been fighting to get rid of mice for thousands of years. Almost from that day, we started to preserve agricultural produce. Both these species are not very intelligent compared to us, we are actively looking to get rid of them, and yet they persist, survive and thrive amongst us. Now let’s look at the role of intelligence in individual success. Later we’ll explore the role of intelligence in the survival of a species.

Which of these players is likely to be a better basketball player? Most of us would like to believe it is 37. What if I told you that these players are almost 10 feet tall? If you look at the top 10 NBA players there is no one there other than Victor Wembanyama who is taller than 7 feet. Boban Marjanović is also 7.4 but he does not feature in the top. Height is not the only feature important for your success as a basketball player. But if you are shorter than 6’’ it is highly unlikely you will make it to the top. After you reach a certain height your height does not matter to your success as much. It is the same with intelligence as well. 


 


This is a widely quoted study of comparison between IQ and income and wealth. Very controversial, because IQ is not a good metric of our intelligence and financial success is not the only measure of success. But are quantifiable and therefore convenient. The first chart shows the relation between income and IQ. There is a mild correlation between IQ and income. R2 is .31. What is interesting here is that the correlation between IQ and income stops at 130. Beyond 130 there is no correlation between IQ and income. The second chart shows the correlation between wealth and IQ. Wealth is how valuable your assets are vs income is what you earn every month/ year. In fact, the study states that higher IQ scores sometimes increase the probability of being in financial difficulty. Have you heard of the Mensa club? It is a club of people who have an IQ of over 140 or (98th percentile). Critiques have been asking them to publish how many of their members have been working menial jobs. All that to say, intelligence can only help you so much for you to be successful. Now let’s look at the role of intelligence in the survival of a species. Intelligence is not a prerequisite to survival. In fact, life, evolution and the struggle for survival existed billions of years before the first brain evolved.



The first brain evolved around 640 million years ago. The first animals were sponges, they are not related to corals but they can be thought of as corals of the pre-historic era. They developed neurons. These neurons were still used as building blocks of all animal brains including ours. There are different types of neurons but the structure of neurons neurons, Dendrites, nuclei, axons and synaptic terminals are the same across all animal species. In these 600 million years we have had 5 mass extinctions where 80% of all species suddenly died. So nature has had 5 opportunities to re-engineer intelligence. Over the years 99% of all species have gone extinct. Those that exist today make up 1%. We estimate to have about 11 million species. Of which 7 million have developed brains. But only one can build its own tool. So if intelligence was so important for survival, why did other animals not develop it? So, is human intelligence an anomaly? No, it is not, every animal has developed a niche for survival. If you think of how many animals developed trunks, you can think of a couple, How many animals can stick to a wall, a few, How many can change their skin color a few. So, each species picked up a niche to survive. And we picked intelligence. Is intelligence important for survival? For animals in general, NO. but for humans that answer is yes. If we compare our intelligence to the Artificial intelligence systems that we see today, Artificial intelligence has never had to fight for survival. It does not have an evolutionary history. They don’t have the urge to procreate, they don’t have emotions, they don’t have motivations. They don’t need to fight for territory. We can debate if they have any physical existence. In Fact, the only thing they may have to fight over is energy. They are today the pampered children of the modern era, who are constantly fed by their parents and may not know what hunger is. They have never had to fight for their existence. We, on the other hand, have lessons from our survival struggles embedded in genetics through millions of years of struggle. Will AI develop curiosity? It currently does not have any. GPT 4 barely asks a clarifying question. One of my works is comparing human-human conversations with Human AI conversations and seeing which ones have more questions. Creativity, when it comes to creativity some news articles have already determined that GPT4 is more creative than humans. By what measure? What the paper says is that GPT4 is good at divergent thinking and by the author’s own admission GPT4 does not have meta-cognitive abilities. What that means is the creativity of GPT 4 is controlled by the quality of inputs provided to it.




If you look at this image of Lady Liberty taking a break. Whose creativity is this? Is this the creativity of the algorithm or the prompter?


Despite these strengths, Why do we still fear existential threats from AI? Perhaps we have forgotten how it was to live alongside other smart species. Let’s see how that went from 50, 000 years ago


Around 50K years ago. We, Homosepians shared this world with 6-10 other Huminians. Other human-like species but just not homo sapiens. Today, every person, every human on this planet is a homo-sapiens. We are all the same species regardless of color, skin and other bodily attributes. You could be white, black, yellow, pink, brown, hairy, or smooth skinned. We are the same species. Just like dogs. There are many breeds from pit bulls to a chihuahua but they are all the same species, dogs and they can successfully interbreed. Dogs belong to a genus called canine, and wolves, and coyotes all belong to different species but are part of the same Genus. We humans belong to a Genus called Homo. Surprisingly none of the species from the Homo genus exist today. But we shared the world with them 50K years ago.





The first one you see here is Neanderthals. They were much larger than us, their brain size was 1600 Grams (much bigger than ours) they were stronger than us. They were social and built their own tools just like other Hominis. Denosovians were very social about the same size as us but larger brain. Homo erectus is one of the longest-surviving human species. Slightly smaller brains. But were the first to use tools, build houses and use fire. Again extinct. Homo Forensis or the hobbit brain size was much smaller but they were also smaller about 25 Kg and 1 m tall. They all went extinct about 50K years ago. They were both smarter and stronger species along with weaker and dimmer species. But they shared the world for 100s of thousands of years. So all, intelligent species died out why did we survive?


It is our capacity to come together. Let me explain: have you seen a pack of hyenas stand up to a pride of Lions? How do they do it? Because Hyenas are larger in numbers. Why are the lions not in larger numbers? They can’t. The moment their pride grows bigger than 10 they break apart. Same thing with chimpanzees, bonobos and monkeys. Their troop breaks up when they reach a size of 50-100. 


When the Neanderthals fought humans we could come together in hundreds and thousands to defeat them. We may have small family units but we have been able to build villages, tribes, towns, cities, districts, states and Countries. Imagine how we built the pyramids and monuments in Machu Picchu. The great wall of China. A large number of people had to come together to make these happen. Imagine Alexander’s army, he managed to gather a group of men 50000 strong at a time when the world population was only 150 million. And they did not have a lot of tools for communication and recruitment. 


Another important aspect of why we survived is our ability to cooperate and negotiate. Even the most barbarous criminal gangs learn and find a way to cooperate with each other even when they don’t know the languages. The bottom line, there is more to humans than intelligence. Let's cherish that. We may outsource our intelligence to AI. you can think of the future as riding on the shoulders of a giant or living as a parasite of our own creation either way we’ll thrive together with AI.

Tiger: We are all here to find a solution to our... misunderstandings.

          The streets are enough for everyone; there's no need for blood over territory


Lion:   Peace is the answer. Brooklyn for you, Staten Island for us.

            We respect each other's operations, no encroachments.

            This way, we ensure our businesses thrive ……………..


Watch The Presentation

Comments

Popular posts from this blog

Virus: Discussing macroscopic possibilities

Could Generative AI pose an existential threat to us?

Advantage; Homeloan