AI: Hype or Hope? Asks Shubhranshu Singh in this week's Simply Speaking
Clay Shirley said “Revolution doesn't happen when society adopts new tools, it happens when society adopts new behaviors”. I think it applies to this case a well," writes Shubhranshu Singh in his Storyboard18 column Simply Speaking.
ADVERTISEMENT
Nvidia announced a booming revenue report a few days ago and now looks set to become the most valued enterprise on earth.
Meanwhile, Elon Musk pulled in $ 6 billion and some change for his venture and his claim is that human-level artificial intelligence is coming next year. Elon Musk’s AI startup xAI has completed this striking fundraising achievement at an impressive $24 billion valuation. Backed by Sequoia Capital and Andreessen Horowitz, xAI is poised to potentially vie with AI giant OpenAI for industry dominance.
This is not his first foray into AI since he co-founded Openai, then left after a rupture with Sam Altman. In his view, Tesla is an AI company not an automobile maker.
Big tech can't seem to mouth enough Al-powered hope statements. Everywhere confidence trumps competence. The more the hype the more the eventual harm.
It sure seems like the ‘Al hype transcontinental flight’ has taken off the long runway whereas in fact it may be already preparing for its eventual descent.
We need only focus on what AI can do, how well can it do these things and productivity will boost returns? AI ventures must generate returns for stakeholders over the medium term.
Several glitchy introductions touted as improvements /innovations are revealing that the rate of improvement for Als is slowing. It naturally makes new ,hitherto unimagined, applications rarer still. Building and running AI on scale is a capital guzzler. What is more , competing Al models are coming into the market continuously whereas the earlier ones are not turning any profits.
On the one hand base AI is getting commoditized but on the other hand staying ahead required spending.
OpenAI's ChatGPT and Google's Gemini-need more data than any steam engine burnt coal. It has to be fed every moment.
Having trained on more or less the entire internet, there isn’t much left.
The free, open-source models, like those from Meta and Mistral, are nearly as good as the current leaders.
If the game shifts from innovative add ons to base cost management, it will need scale of subscription to remain afloat.
Al startups have already started to go belly up or selling out to the leaders of the pack. The Wall Street Journal reported on the exit of the Inflection AI’s team to Microsoft whereas the CEO of Stability AI cut away in March.
A report in the Wall Street Journal quoted Sequoia’s estimate of $50 billion dollar having been spent on chips from Nvidia to train Al in 2023. But the revenue for all players is only $3 billion in revenues.
Google, now offering AI-generated summaries across billions of search results, will eat into its margins since redirection and programmatic advertising is its main source of revenue.
Clay Shirley said “Revolution doesn't happen when society adopts new tools, it happens when society adopts new behaviors”. I think it applies to this case a well.
How many corporations have put Al at work in their mainstream functions?
How many license and pay for Al tools for their workforce at large?
Being intrigued by Al and bearing a cost to use it and add value are two very different things.
Financial Times reported in December that OpenAI’s revenue was at least $2 billion, and would double by 2025.
But OpenAI has nearly $90 billion market valuation and at some stage the multiples will look hyper optimistic. This is the case for Anthropic having raised 8 billion dollars and being valued at 18 billion.
Is Al a tool or master? Is it an enhancer of productivity or -in certain classes of jobs- entirely a substitute for human effort?
Training people and building habits /workflows remain barriers to the swift adoption of Al.
Al may well be radically disruptive . It may well transform jobs and human lives.
But at its current pace, it is likely to take more fuel and cover less distance. Hence, its viability is a question. A great tool is that which has the least variance of performance.
I would say it’s hype over hope for now.
Shubhranshu Singh is VP and CMO, Tata Motors CVBU. He writes Simply Speaking, a special column on Storyboard18. Views expressed are personal.
