Flaky_Suit_8665•This week
[D] The current and future state of AI/ML is shockingly demoralizing with little hope of redemption
I recently encountered the PaLM (Scaling Language Modeling with Pathways) paper from Google Research and it opened up a can of worms of ideas I’ve felt I’ve intuitively had for a while, but have been unable to express – and I know I can’t be the only one. Sometimes I wonder what the original pioneers of AI – Turing, Neumann, McCarthy, etc. – would think if they could see the state of AI that we’ve gotten ourselves into. 67 authors, 83 pages, 540B parameters in a model, the internals of which no one can say they comprehend with a straight face, 6144 TPUs in a commercial lab that no one has access to, on a rig that no one can afford, trained on a volume of data that a human couldn’t process in a lifetime, 1 page on ethics with the same ideas that have been rehashed over and over elsewhere with no attempt at a solution – bias, racism, malicious use, etc. – for purposes that who asked for?
When I started my career as an AI/ML research engineer 2016, I was most interested in two types of tasks – 1.) those that most humans could do but that would universally be considered tedious and non-scalable. I’m talking image classification, sentiment analysis, even document summarization, etc. 2.) tasks that humans lack the capacity to perform as well as computers for various reasons – forecasting, risk analysis, game playing, and so forth. I still love my career, and I try to only work on projects in these areas, but it’s getting harder and harder.
This is because, somewhere along the way, it became popular and unquestionably acceptable to push AI into domains that were originally uniquely human, those areas that sit at the top of Maslows’s hierarchy of needs in terms of self-actualization – art, music, writing, singing, programming, and so forth. These areas of endeavor have negative logarithmic ability curves – the vast majority of people cannot do them well at all, about 10% can do them decently, and 1% or less can do them extraordinarily. The little discussed problem with AI-generation is that, without extreme deterrence, we will sacrifice human achievement at the top percentile in the name of lowering the bar for a larger volume of people, until the AI ability range is the norm. This is because relative to humans, AI is cheap, fast, and infinite, to the extent that investments in human achievement will be watered down at the societal, educational, and individual level with each passing year. And unlike AI gameplay which superseded humans decades ago, we won’t be able to just disqualify the machines and continue to play as if they didn’t exist.
Almost everywhere I go, even this forum, I encounter almost universal deference given to current SOTA AI generation systems like GPT-3, CODEX, DALL-E, etc., with almost no one extending their implications to its logical conclusion, which is long-term convergence to the mean, to mediocrity, in the fields they claim to address or even enhance. If you’re an artist or writer and you’re using DALL-E or GPT-3 to “enhance” your work, or if you’re a programmer saying, “GitHub Co-Pilot makes me a better programmer?”, then how could you possibly know? You’ve disrupted and bypassed your own creative process, which is thoughts -> (optionally words) -> actions -> feedback -> repeat, and instead seeded your canvas with ideas from a machine, the provenance of which you can’t understand, nor can the machine reliably explain. And the more you do this, the more you make your creative processes dependent on said machine, until you must question whether or not you could work at the same level without it.
When I was a college student, I often dabbled with weed, LSD, and mushrooms, and for a while, I thought the ideas I was having while under the influence were revolutionary and groundbreaking – that is until took it upon myself to actually start writing down those ideas and then reviewing them while sober, when I realized they weren’t that special at all. What I eventually determined is that, under the influence, it was impossible for me to accurately evaluate the drug-induced ideas I was having because the influencing agent the generates the ideas themselves was disrupting the same frame of reference that is responsible evaluating said ideas. This is the same principle of – if you took a pill and it made you stupider, would even know it? I believe that, especially over the long-term timeframe that crosses generations, there’s significant risk that current AI-generation developments produces a similar effect on humanity, and we mostly won’t even realize it has happened, much like a frog in boiling water. If you have children like I do, how can you be aware of the the current SOTA in these areas, project that 20 to 30 years, and then and tell them with a straight face that it is worth them pursuing their talent in art, writing, or music? How can you be honest and still say that widespread implementation of auto-correction hasn’t made you and others worse and worse at spelling over the years (a task that even I believe most would agree is tedious and worth automating).
Furthermore, I’ve yet to set anyone discuss the train – generate – train - generate feedback loop that long-term application of AI-generation systems imply. The first generations of these models were trained on wide swaths of web data generated by humans, but if these systems are permitted to continually spit out content without restriction or verification, especially to the extent that it reduces or eliminates development and investment in human talent over the long term, then what happens to the 4th or 5th generation of models? Eventually we encounter this situation where the AI is being trained almost exclusively on AI-generated content, and therefore with each generation, it settles more and more into the mean and mediocrity with no way out using current methods. By the time that happens, what will we have lost in terms of the creative capacity of people, and will we be able to get it back?
By relentlessly pursuing this direction so enthusiastically, I’m convinced that we as AI/ML developers, companies, and nations are past the point of no return, and it mostly comes down the investments in time and money that we’ve made, as well as a prisoner’s dilemma with our competitors. As a society though, this direction we’ve chosen for short-term gains will almost certainly make humanity worse off, mostly for those who are powerless to do anything about it – our children, our grandchildren, and generations to come.
If you’re an AI researcher or a data scientist like myself, how do you turn things back for yourself when you’ve spent years on years building your career in this direction? You’re likely making near or north of $200k annually TC and have a family to support, and so it’s too late, no matter how you feel about the direction the field has gone. If you’re a company, how do you standby and let your competitors aggressively push their AutoML solutions into more and more markets without putting out your own? Moreover, if you’re a manager or thought leader in this field like Jeff Dean how do you justify to your own boss and your shareholders your team’s billions of dollars in AI investment while simultaneously balancing ethical concerns? You can’t – the only answer is bigger and bigger models, more and more applications, more and more data, and more and more automation, and then automating that even further. If you’re a country like the US, how do responsibly develop AI while your competitors like China single-mindedly push full steam ahead without an iota of ethical concern to replace you in numerous areas in global power dynamics? Once again, failing to compete would be pre-emptively admitting defeat.
Even assuming that none of what I’ve described here happens to such an extent, how are so few people not taking this seriously and discounting this possibility? If everything I’m saying is fear-mongering and non-sense, then I’d be interested in hearing what you think human-AI co-existence looks like in 20 to 30 years and why it isn’t as demoralizing as I’ve made it out to be.
​
EDIT: Day after posting this -- this post took off way more than I expected. Even if I received 20 - 25 comments, I would have considered that a success, but this went much further. Thank you to each one of you that has read this post, even more so if you left a comment, and triply so for those who gave awards! I've read almost every comment that has come in (even the troll ones), and am truly grateful for each one, including those in sharp disagreement. I've learned much more from this discussion with the sub than I could have imagined on this topic, from so many perspectives. While I will try to reply as many comments as I can, the sheer comment volume combined with limited free time between work and family unfortunately means that there are many that I likely won't be able to get to. That will invariably include some that I would love respond to under the assumption of infinite time, but I will do my best, even if the latency stretches into days. Thank you all once again!
![[D] The current and future state of AI/ML is shockingly demoralizing with little hope of redemption](https://jfbmhhfxbbrxcmwilqxt.supabase.co/storage/v1/object/public/resource-images/MachineLearning_AI_implementation_20250328_185811_processed_image.jpg)
