AI Death or Dud?
As we debate if AI will kill us or be a passing fad, Big Tech positions for astronomical wins
Happy Friday all,
Loads of negative headlines recently about whether Generative AI is a ‘dud.’
The whole AI debate seems to oscillate between extremes. Either AI is so powerful that it’s going to kill us all, or it’s a fad and means nothing at all.
It can’t be both, can it?
Personally, I think the focus on the existential risk is greatly exaggerated. Not because it isn’t hypothetically possible — it is. However, the overwhelming focus on this possibility distorts the current reality. Mainly because it:
Gives AI undue power — as if machines were autonomously plotting the future of humanity.
Takes away human agency — as if humans were not, and will not continue to be, a central part of AI development.
Paints AI development as a net negative — as if the pursuit of machine intelligence is analogous to the development of atomic weapons. (This reminds me of Robert Oppenheimer's reflection as he witnessed the first detonation of a nuclear weapon in 1945: "Now I am become Death, the destroyer of worlds.” I’ll write more extensively on this soon.)
Nonetheless, the current narrative around super-intelligent machines playing Russian roulette with our mortality is getting hackneyed. And so the pendulum is swinging back — all the way back.
Now, Generative AI is apparently a dud. Well, that’s not right either.
Generative AI: Dud or Death?
A lot of this school of thought has been focused on the money.
Undoubtedly, much of the money flowing into the space will be lost. As we’ve looked at previously, it’s ‘silly money’. But while everyone and their dog is starting a ‘Generative AI’ business, only a few will really withstand the test of time.
Indeed, much of the GenAI investment — though at record highs — is centralised around a few key players.
Even then, some of them already look shaky. We know that Stability AI, a company that played a seminal role in the story of GenAI with their open-source release of Stable Diffusion last year, has reportedly been having trouble raising money.
Additionally, there’s been a lot of negative press about how many freshly minted Unicorns in the space don’t have sustainable business models.
But that is NOT what the data suggests.
If anything, we’re starting to see the business models for GenAI — especially for AI infrastructure and enterprise applications of models such as large language models — materialise very quickly.
This week, we found out that OpenAI, the poster child for GenAI, is on track to generate a billion dollars in revenue by 2024…! Not so bad for a company that lost 540 million dollars developing the GPT-series last year.
And it’s not only OpenAI.
More records are being broken by Nvidia too. On Tuesday, its shares rose 4.2% to close at a record high after announcing its partnership with Google to build super-powered AI infrastructure.
So far, NVIDIA’s stock is up 234% in 2023, making it the best performer in the S&P 500. Coming in second? Meta, which is up 148%. As previously discussed, Meta’s stunning comeback has everything to do with AI.
The Normalization of GenAI
To a certain extent, I think the ‘dud-lash’ regarding Generative AI has to do with the maturing of our understanding— and the normalization — of the technology and its applications.
The ‘ChatGPT moment’ was so awe-inspiring because, only nine months ago, GenAI’s capabilities seemed almost omnipotent. As we all know, the birth of the ChatBot riggered endless (ongoing) discussions about AGI and mass extinction.
Now, some of us are already over it - way over it.
So perhaps we can draw two main conclusions from why we are already debating if Generative AI is a dud.
First, we adapt to the new reality quickly.
We’re no longer impressed by a conversational LLM or an image-generating AI model. For those who work with or use AI, these capabilities are already part of what is expected in a digital workflow. Now, we want to optimize. If AI can ‘generate,’ we need to generate exactly what we need for our unique use case.
Second, working with AI requires … work.
Sadly, AI is not a magical and autonomous agent —yet. Getting an AI system to deliver real value for any intelligent or creative endeavour requires a heavy lift. We need to figure out how to prompt it, train it, and to generally mould it to do our bidding. No shortcuts!
A New Frontier
But while the public debate sways between the extremes of ‘AI will kill us’ and ‘AI is a fad’ (a narrative I expect will get louder in the months to come), another reality is quietly taking shape.
GenAI represents a new frontier in the quest for intelligence. While the current phase is marked by the deployment of enterprise applications for models like LLMs — this is just the tip of the iceberg.
But the tech giants, those best positioned to capitalize on the Intelligence Revolution, are moving at a breakneck pace. They are building, refining, and investing — as they vy to control the essential infrastructure underpinning this Revolution.
So, while we oscillate between debating whether AI will annihilate us or if it's just a flash in the pan, they are steadily working towards owning it.
We should take note.
Namaste for now,
Nina