Introduction to the Concept of Artificial Intelligence
The recent corporate fiasco that ridiculed the board of OpenAI, the unicorn behind the famous ChatGPT, gave rise to the craziest interpretations.
Reuter's recently circulated a rumor claaiming that OpenAI engineers achieved a major milestone by developing an AGI (Artificial General Intelligence) named Q* (Q Star). This AGI supposedly possesses autonomous reasoning and decision-making capabilities. However, the field of AI is serious business, and there's an urgent need to inform the public, especially in the context of investments. To get insights on this, we'll turn to the Incomplete Singularity expert, Nissam Gorma.
“Artificial Intelligence”, the word has been on everybody’s lips since the ChatGPT boom. Let's be honest: as ChatGPT is the first AI-based tool with which the public can easily access, i.e. requiring no effort of research, training, or prior analysis to interact with it; it mainly drives ROI-enhancement fantasies of 99.9% of the finance industry, whose frenzy lazily pivots according to the latest trendy tech concept. Exit blockchain, exit Metaverse, exit Web3: brace for the landing of AI!
You are launching a startup focusing on an AI skateboard, an AI toaster, or an AI refrigerator? You will raise your seed round without any difficulty! Are you setting up a VC fund focusing on AI skateboards, toasters, and refrigerators? You too will have LPs (Limited Partners) running to you to invest in no time! Very few have seriously studied the subject from a mathematical, algorithmic, or academic point of view. Very few know what AI really is, but everyone wants to invest!
There are two categories of investors
Private investors, professional or institutional, small or large... In reality, there are only two kinds of investors in tech - I am voluntarily limiting myself to a sector that I have been navigating for many years - and my feeling is that it is already one too many categories: there are those who work and those who don't. Oddly, it is often “the other one” who we hear moaning about its losses and demanding compensation when the first scams collapse. But the other side of the story is: as long as it bites, they pull the line - especially in the US startup ecosystem. Money, more money, and even more money… until resounding scandals, damages costing billions of dollars, as in the case of Theranos in biotech, and more recently with FTX in crypto-assets, to name only two.
In this context, “work” therefore means: doing the research and analysis required to be able to make investment decisions with rock-solid conviction supported by logical and numerical reasoning, and to acquire the ability to understand with sufficient depth the fundamentals of the underlying sector. Doing the job requires a few thousand hours (at minimum) spent reading, discussing, traveling to trade shows, or attending specialized conferences. To continue with the full disclosure we have been setting up with this piece: asking interns to do the research and producing a 6-bullet-points summary will only be proven useful in bars – and will be far from having done the job.
Therefore, how can we educate people on a scientific subject in constant technological evolution, linked to another highly complex technological disruption (quantum computing), which requires dozens of hours of evangelization? “World of the Machines” is a new project launching this week that aims at engaging with the general public in a fun way about AI and its consequences. We will come back to this in detail in the coming weeks. Let’s stick to the prerequisites first.
The Halo Effect
Tim Urban brilliantly analyzes the halo effect to which the terms Artificial Intelligence and Singularity are victims: practically no one evaluates them from their academic definitions, but rather from the chaotic premises of their concrete and material reality, or from the fanciful interpretation that is made of them in the entertainment industry (e.g. films, SF novels). The judgment we have of these terms mainly comes from the symbolic halo that they radiate. These terms are therefore ultimately never used for the definition they carry but overused even before becoming a reality. Thus, it is difficult to hope for engineering correct reasoning on these subjects…
Artificial Intelligence - when “the work is not done” - is Terminator, Hal in 2001: A Space Odyssey, more recently the android Ava in Ex Machina, or the hologram Joi in Blade Runner 2049. It is also the “neural engine” introduced by Apple in its series of type A chips (starting with the A15 Bionic in the iPhone 13 when it was released in 2021), or the autopilot of Tesla cars (based on algorithms such as Convolutional Neural Networks, Recurrent Neural Networks, and Reinforcement Learning).
Singularity is the entity in Mission Impossible – Dead Reckoning, Alphie in The Creator, or more recently Q* in the “secret part” of the OpenAI labs.
It is difficult to understand what AI is actually capable of in real life. But also, does the Singularity, this human brain reproduced in a program running on a silicon chip already (or will) exist?
- Do not confuse fiction with reality
- Have the most relevant assessment possible with current accessible knowledge, and separate manipulated concepts from their exact definition
- Educate yourself and grow your discernment on these subjects
That’s “doing the job”. Investing even one cent in this sector without having overcome this threshold would simply be irresponsible.
AI: What is it really?
John McCarthy, who coined the term “Artificial Intelligence” in 1956, complained that “as soon as it works, no one calls it AI anymore.” Which makes the term “AI” seem more like a mythical prediction of the future than a concept destined to become a reality. The term also often evokes a pop concept from the second half of the 20th century which never came to life. Ray Kurzweil explains that he often hears that AI "ran out of steam" in the 1980s. A phenomenon he compares to the claim that the Internet was dead during the collapse of Internet companies in the early 2000s (the dot-com bubble).
Two decades later, we know what the Internet has become: we can therefore have an idea of what AI can become. Again, AI is really no joke.
To clarify things between the initial definition of 1956, the definition revised in the light of current advances, and the “ideal” definition of what the concept could become in the future, experts had to create a cluster of subcategories falling within the framework of what we call Artificial Intelligence.
Among them are:
- ANI (Artificial Narrow Intelligence) which would define the current technological state;
- AGI (Artificial General Intelligence) which would define the state relayed by the rumor concerning Q* escaping from OpenAI labs; and
- the feared but futuristic ASI (Artificial SuperIntelligence), capable of taking control of everything connected, from nuclear missile launch silos to the doors of high-security prisons and the entire global banking system, or the access to the global energy grid as well as industrial, military and medical infrastructures…
We understand that it starts getting complicated.
Singularity: What is it really?
This term was initially used in mathematics and physics to designate a system evolving towards a state of infinitely small, infinitely dense, infinitely hot, in short, a “limit” state that no longer responds to fundamental laws. It was in 1993 that it was coined by Vernor Vinge in his eponymous essay (The Coming Technological Singularity) and then defined the "moment in the history of humanity from which intelligence (as a capacity for reasoning and decision relevance) of our technology will become superior to ours.” Subsequently, the term will be confused with the moment in history from which such intelligence appears, and after that simply as this kind of intelligence - provided that it is unique because nothing allows us to affirm today that only one Singularity would prevail…
Furthermore, it is important to note that it has never been fixed that this intelligence would be based on ours. Indeed, it would logically suffer from the same defects and have difficulties overcoming ours. It is also probable that we do not have the capacity to understand the way in which such intelligence reasons: at most, we could understand or simply try to interpret its effects (what it affirms, what it decides).
Finally, it has never been fixed either that this intelligence necessarily resides in a silicon-type material, or that it uses transistors.
But then, what are we talking about?
Just imagine that with the current knowledge about AI and Singularity, I only mention the information, publications, and research papers available to the general public. The situation would obviously be more nuanced (to say the least) should we have access to the highly classified military AI labs located somewhere in Asia and in the US; the equivalent to P4-type laboratories researching the deadliest viruses there are, and from which the Covid would have escaped according to the rumors (again)…
“Doing the job” thus also means watching the latest videos from Boston Dynamics, then noticing that it was acquired by Google in 2013, then quickly taken over by SoftBank, but that it is now in the hands of Hyundai WIA, a corporation specializing in the design of ultra-sophisticated weapons, a subsidiary of Hyundai Motor Group, the automotive manufacturer well-known to the general public.
To better understand what we're talking about, to help you "do the job" and not take at face value the announcements or rumors that will be accumulating over the coming weeks and months about AI and the emergence of one or more Singularity from the labs of this or that startup in need of 10-figure funding, I invite you after this first introductory part to follow the presentation of “World of The Machines”. The project launched this week, and its main protagonist is the “Incomplete Singularity” named Nissam Gorma. Through this fascinating “gamified” artistic and technological experience based on a unique connected fidget (for finger-gadget), the founding team aims to challenge the general public about what really is Artificial Intelligence, and most importantly what it will probably be capable of in a few years when it actually reaches the Singularity stage.
Come back next week for part 2!