For the last few weeks I’ve been talking about the importance of gaining solid scientific intuition about AI in order to have a better grasp on our individual place in this fast paced space. The space that is quite promising yet over-hyped, confusing and at times scary. I alluded to the AI ethical implications that everyone, particularly AI researchers, should bear in mind when stepping into uncharted waters of bringing to the world yet another “biggest-bestest-mightiest” LLM.
In this and the upcoming articles I continue exploration of:
- what is the current AI landscape and its driving forces,
- what is our individual role in it and what should be our attitudes about it,
- how shall we adjust to the new realities and be better prepared for the uncertain future.
State of the AI Union
It seems bizarre how Silicon Valley by and large was totally unprepared to usher into the new era of AI prominence in a civilized manner. The “ugly” existential human instincts got exposed in the course of a very short time over the past a year plus. Let’s visualize what are we dealing with.
Hooligan
OpenAI went unhinged by hastily exposing to the public a half-baked product, ChatGPT, at the end of 2022. It was destined for a perpetual beta-testing while potentially becoming a very much impactful to humanity tool. A totally opportunistic tsunami wave moment.
Unlike Microsoft or Google, OpenAI didn’t have legacy products to worry about, nor infrastructure. They didn’t have anything to lose. They are all-in in this race for AI dominance.
Opportunist
Microsoft, audacious bettor and financial backer of OpenAI sudden move, seizes an opportunity to come back from murky relevance in the marketplace and reinvigorates its not much exciting anymore productivity tools with AI spicing.
In addition they have infrastructure – Azure – which enjoys much needed revitalization in hot cloud business. The more AI models are trained and deployed in production, the more money their datacenter furnaces will bring to them.
Confused Monopolist
Google, a recognized factory of A-class brains behind the AI and Machine Learning breakthroughs in the past decade, was caught by surprise. It felt its monopolistic dominance in internet traffic monetization thanks to smart ads placements is going to slip away.
They are in a really tough situation:
- support aggressively AI and the next thing you see cash cow ads disappear,
- slowing down on AI adoption can render you irrelevant in the not so distant future.
A lot is at stake for Google.
Sneaky Robin Hood
Meta (Facebook) sneaks in by throwing all the internal resources to build a so-called open-source LLM, and portrays itself as a white-night rushing to you with the Llama LLM series to democratize AI access. At least that’s what they claim.
What do they have to lose? Anyway they are sitting on the trove of people behavior analytics. Throw the freebee potent LLM into masses, obliterate the competition. After all, to maintain and upgrade Llamas will have dependency on Meta. You can’t afford a cluster of 24,000 GPUs to upgrade your version of LLM, can you?
(Still-) Undecided
Anthropic split from OpenAI and spearheaded prioritization of human alignment with its Claude. But as time goes by one can’t help to ask exactly how OpenAI and Anthropic differ.
Model Hubs
Amazon, IBM chose to be the hubs of choice catering GenAI models to their respective customer base with a plethora of open-source LLMs.
Hugging Face is a new entrant which became a de-facto independent hub for open-source LLMs. Worth noting that 96% of them are redundant in their capabilities. It is not clear at all how many of the claimed million plus hosted models are of true if any practical value.
Wannabes
Then there are Cohere, Mistral, Databricks and a dozen of other smaller newcomers or prominent incumbent data science players vying to make their dent in this promising yet highly speculative and overcrowded GenAI space.
Winner
They all are still disproportionately spending more than earning from GenAI offerings. There is one single undeniable benefactor in this carnival of AI future promises – that is NVIDIA.
For them the future is here already – they make lots of money by having monopolized GPU space that fuels data center furnaces that are training and operating LLMs. They are your Saudi Arabia fossil fuel of Silicon Valley.
“Victims”
Finally, we have the rest of the world:
Hundreds of vendors stepping on each other’s toes claiming having right cogs for the GenAI machinery – vector databases, orchestration etc. More often than not with very redundant and short shelf-life value propositions.
- Enterprises that are extremely cautious onboarding unproven AI technology. They are in an analysis-paralysis mode. They are nervous. They know the stranger got into town for good and will relegate them to irrelevance soon if they don’t move fast (-enough).
- Many boutique outfits all of sudden became “one-and-only AI Experts” and influencers that handily outnumber your neighborhood bakeries.
- Mid-level career professionals, college graduates are no less nervous – exactly in what shape and form their professional futures and lives will upend.
Silicon Valley is unprepared
With the massive amount of investments and ensuing AI arms race for the land-grab unraveling in such a short period of time, Silicon Valley PR was caught completely off-guard. The results of this didn’t wait, today we witness extreme polarization of society.
We see sides with firmly formed opinions ranging on the spectrum from profusely glorifying AI and calling to drop anything else not supportive of AI expansion to doomsday oracles outright calling for AI ban. Anything in-between filled up with FOMO (fear of missing out) exacerbated by very opportunistic investments or, quite opposite, analysis-paralysis.
Of course, we’ve been in a similar situation before with the advent of technological breakthroughs (e.g. typographical press, electricity, internet to name a few). But none of those monumental advancements were happening so fast and simultaneously penetrating every corner of human’s life with a very vague understanding of social and economic consequences.
Overhyped AI space, polarization in the society, and social and technical complexities associated with AI don’t help stay level-headed. Striking the consensus is extremely hard. Conway’s Law stating that “any organization that designs a system (defined broadly) will produce a design whose structure is a copy of the organization’s communication structure” increasingly applies to AI space too. People take sides, they want to be listened to but don’t bother to listen to another side. In absence of convergence and genuine reach across – it is hard to expect progress with no collateral damage.
So we are left to sort out this conundrum ourselves, on the fly, with the AI locomotive speeding somewhere ahead at 500 mph.
But it is not all lost. I am a firm believer that in given circumstances in order to keep up and not feel lost, the key is our continuous self education. Armed with proper knowledge we will be able to timely adjust and avoid anxiety. For that we need to continuously keep our hand on the pulse:
- understand how to get intimate with AI – it is here to stay, don’t kid yourself,
- find our way to upskill self with AI being our co-intelligence mate,
- align with the players (companies, organizations) who are close to our values and philosophies of life.
With proper education we will prosper. Stay focused on consuming proteins, not empty carbs about AI. We can do better than Silicon Valley!
See you next time on these waves 🙂
Leave a Reply