(1/3) Tackling the future: experts do harm when people believe them?
I recently started following Ethan Evans, his intriguing takes on AI and wise advices on senior management are always a curious read of…
I recently started following Ethan Evans, his intriguing takes on AI and wise advices on senior management are always a curious read of mine. He took part of the TED Conference 2024, a prestigious event I can just wish to be part of ever.
However, an article about Vinod Khosla’s presentation grabbed my attention thoroughly. As a technocrat myself, being in the software engineering business for over a decade and a half, an avid reader and consumer of engineering and entrepreneurial topics, and being an optimist myself, the article contained claims I could not stop thinking about. Read here Ethan’s notes for a quick overview:
Fortunately Vinod shared his slides to the public so the slightly extended content could explain a bit more of his points:
While researching backgrounds and what was actually presented so I can understand the points myself, found only limited reports:
I immensely respect Mr. Khosla’s track record in tech and investment. His mindset of endless optimism, as it is shining through the points he put out, most probably led teams through breakthroughs otherwise hardly conceivable. That said, there must be a check on positive thoughts as they can lead astray to obviously untenable solutions. So let me try to be the sound of reason a bit.
I will go through each point below, sticking to a high level and share my immediate thoughts, the challenges surrounding the “possible future”. And as just criticizing is easy for anyone, I will try to put out at least one immediately implementable idea to advance such causes and direct to sources that are much more informed than me.
“Experts extrapolate the past… and do harm when people believe them”
I understand the need to start off strong in today’s attention deprived world, an inflammatory statement all but guarantees the focus of your audience. I bite on the trap.
This is still an extremely narrow understanding of what an expert is and what actually their role is. So let’s take basic logic, as an entrepreneur, you want to build a product, a service, a solution to a problem that the market values and you get return and scale your business. For this, you need capital and expertise and as humans, we are inherently limited in one resource: time. Time is essential for each human to learn, develop, work, sleep, etc, time limits what you can know and do alone, so we have a great invention, called team-work where specialized individuals, if well organized, are capable of more then any of them would be alone.
Beyond this, I understand that maintaining or incrementally improving infrastructure and existing solutions are not “sexy” for any investment, especially for venture capital that tries to maximize the return by finding so unique solutions, that their worth can pay off a thousand failed experiments. It is an interesting and effective tool, however not the only one in our arsenal.
“Entrepreneurs, with passion for a vision, invent the future they want.”
Discarding expert opinions is a dangerous mindset and will greatly backfire. Past, present and future are all intertwined, you can’t build a future without a foundation existing. That foundation is our present, which in turn was built on top of our past. Learning history, you will see that every event is a logical consequence of the chain of previous events, without experts who understand these events, you will just repeat the mistakes that could have been avoided and waste resources, time and money doing so, that could be utilized to try new ways to succeed.
Putting all your eggs into the utopian visions without a way to get there is foolish at best, destructive at worst. Any one can invent all the futures they want, can even write it, let’s call it science fiction, however if there is no technology, no materials or processes to make it a reality, then it becomes a research. Foundational research inherently risky, as there is no guarantee that you can find a solution, let alone a viable solution and way more likely finding a thousand ways not to reach an impact.
To be clear, these foundational research are necessary for progress, the key is that never there is a guarantee in them so you can hedge your investment taking different chances at the same time.
And now the points
I won’t reiterate all details of each point, please read Mr. Khosla’s presentation on the link above to get his ideas.
1. “Expertise will be free”
This point, assigning low value to “expertise” is a returning theme in these slides. First, I clarify what does expertise mean to me: it is the knowledge of a niche topic, understanding of basic principles and application of those principles to build/develop solutions and to further that field of expertise by having a feedback of successes and failures of each endeavor.
This is not yet possible with AI alone and as a software engineer working around the field (not in it actually intentionally), I don’t think it will be in the next 100 years. Current AI technology is a statistical prediction engine, figuring out what is the most likely next bit of information, be it a word, a musical note or a color of a pixel. These can feel like they have human expertise, but they don’t. They routinely imagine things as aptly called “hallucinating” instead of the good old “it’s a bug” phrase, and I could imagine how devastating an AI doctor’s hallucination could be for the patient it cares for. Photo-generation tools routinely imagine a sixth finger on a human, I really don’t want to argue with my personal AI doc that my stomach ache does not go away because I am missing a sixth finger on my left hand…
In contrast, I believe actual expertise will be even more valuable and expensive to the level that it can become prohibitively costly. Look at healthcare systems around the world, most experience shortage of trained and experienced staff and caretakers already.
What could be done now, is to increase research literacy by subsidized programs, enabling specialized LLMs to be used as assistants for doctors and nurses and feed it the patient’s history with context data from the neighborhood. This itself would increase digitalization in hospitals and clinics, with access to cheap smart phones and the ubiquitous Internet. This would require national governments and private sector service providers to work together securing the patient’s sensitive personal data.
Focused only on healthcare and we are not much further on education either, today even decade old techniques are not widely spread, modern technology is barely utilized in classrooms and even that has endless applications and benefits. Just think of kids learning team-work, problem solving and all around be a professional workplace force just by playing Minecraft under a well equipped teacher. Imagining such as a parent, it would be much harder to fix an AI’s hallucination of fictitious world history than to just sit down with my (future potential) kid and learn it correctly for the first try. And do we want the next generation to be even more separated from human interactions in their formative early years?
2. “Labor will be near free”
A few companies achieved great results lately with bipedal and various other robots simulating work, the chef and the warehouse worker videos come to my mind for this. This is the first point I mostly agree with Mr. Khosla, that the robot industry will become even larger at one point than the automotive, especially if we combine it with the efforts to go to space, where any robotic assistance would be extremely important as they conserve otherwise limited resources like air and water, using more abundant such as solar and nuclear energy.
However this point is less optimistic or utopian if you think about it. First of all, vast amount of employees will be terminated as their physical work becomes redundant or even a bottleneck in the robot-run factory. Without government regulations catching up in time, introducing universal basic income, the situation can become a humanitarian crisis. Actually the Expanse sci-fi series predicted this possible future in my opinion the best: most of Earth’s residents do not work as simply there is nothing to do. Many wait for decades and become old before they can get their first job or even education by lottery.
What could be done now: we have to prepare the gradual, later disruptive changes in workplace supply-demand, upscale or side-scale workforces by subsidized trainings toward skills that are in short supply, maintenance and repairmen, electricians will be a huge need just to keep on the robot industry.
3. “Computer use will grow expansively”
A lot to unpack here, from conflating or undervaluing programming skills, how programming languages are tiered, to contradicting the previous point about robots, etc.
I would like to address at least one aspect, we currently have many solutions to bring “programming” closer to non-engineers, from drag and drop designers (e.g. Unreal’s blueprint), to intent-based languages (like SQL’s syntax) and low-code solutions. In reality, each of them trade in pros and cons based on the given problem and available resources to solve it. If you need a small, 1–2 page UI to edit a few data points, you can do it in these solutions, relatively quickly and without specialized training/experience.
However, even if you scale the problem a tiny bit up, like you need to do a few dozens of data points on a screen, or you need to process banking transactions on time limit and scale, “programming” becomes the least of the concerns and the experience of solving problems, scaling complex systems and the knowledge how to avoid the pitfalls and waste enough resources to bankrupt yourself while trying to solve the problem skyrocket in importance.
Understanding the problem and constructing an effective and efficient solution are the main challenges with software, not writing the code itself. Additionally, generated code, even AI generated code, are a nightmare to debug, fix, and understand, and would probably throw out and do it myself in less time.
What could be done now: focus training in evergreen skills and abilities, like problem solving, team-work, story-telling and influencing, public speaking techniques and combine soft skills with practical experience so the actual software engineers don’t have to clean up the billion+ would-be programmers generating way-less-than-ideal code.
4. “AI will play a large role in entertainment”
Beyond the gray area of copyright and intellectual property rights now composed with the ever-larger-data-set eating machine learning models, this point creates another divide inside me.
On one side, it would be awesome to get personalized entertainment, in the same style of my beloved sci-fi series I grew up on. On the opposite side, it may cheapen the originals to the level that in this new world, they might not have ever been made. Art, which includes television, cinema and gaming, is to tell an idea or story, to bring out emotions, to make you think and it needs an author’s purpose behind to worth it, today’s AI does not have its own purpose or wants.
And then the secondary issue, let’s imagine it works and now there are a large amount of AI generated personalized whole night movies and binge-able series are out there, shared, rehashed again and again, consumed and then generated again based on hearsay.
And now what content these AIs train on? Those generated by other AIs, then again, and again, until the content is so distorted, the style is now unrecognizable and human touch, an artistic depiction of a dilemma and choices, of characters acting and deciding, conflicts and climaxes of stories are just a series of guess-work from an algorithm. I can hardly imagine this worth it if there is no intent behind to tell a story. Though I could imagine a Black mirror episode easily (don’t spoil please, I could not watch all of them yet!).
And authors, celebrities and consumers will drift further apart from each other, into their personally customized worlds, their own realities, where actually anyone can be a celebrity. “AI, make me a large crowd of followers, sprinkle it with journalists and vloggers wanting to make an interview of me.” Of course it won’t be real, but then at this point, what could be?
What can be done now: first of all legislation have to catch up to the now and handle the age of Internet appropriately, then it can look forward especially in the area of copyright, authors and distribution rights. Artists now could learn to integrate AI tools into their workflows with the ethical limits respected, or outright choose to avoid such tools for the sake of an elevated authenticity of their projects.
(E.g. I totally avoid using any AI, even for ideation and research for my weekly sci-fi series so I learn more of writing and more of my personality comes through the pieces. Although I understand many authors want to use the new tools to their benefits.)
To be continued soon with points 5 through 8 in next article.
Thank you sincerely
Thanks to Vinod Khosla giving these points, inspired to think forward and I am sure this was his main purpose, to spark curiosity combined with hope for a better future.
Thanks for Ethan Evans to spreading it to his audience as it would not be on my radar otherwise.
And thank you my dear reader for your time and attention! If you enjoyed, leave me a clap or two! If you have any thoughts to share, challenges or comments, please write on the article, looking forward to read your thoughts!