AI MODELS – MIDDLE LAYER TO MOONSHOTS
Artificial Intelligence, AI, has been the mega buzz word for quite some time. What is being attempted or being talked about may not be even on fringe of it, but the garb of AI for everything for the tech trendy guys seems to have become the trend. If you not in AI you don’t belong to tech space seem to be the statement of the day. But the serious tech world knows for sure, where does the real work – research is happening and it is only from there the news and models come. OpenAI and Google DeepMind have made a name for themselves. When AI enabled natural language processing models are making news, the futurist in you, looks forward to it.
OpenAI spearheads the natural language research models world. With CoPilot and DALL-E-2 under their belt they certainly have come a long way. GitHub CoPilot has also been keeping up pace. Undeniably, the products made and being made are likely to bring seismic changes in the tech landscape and consequently, in the life of mankind. How it would unfold or for even a small glimpse of the trajectory, there would be no better person than Sam Altman, the CEO, of OpenAI. To say they are redefining the future of mankind is to state the obvious.
In a recent interview he has given some pointers. The well accepted thought processes in this field have been thrown to the winds. The readymade mindset will not deliver is the message. He states quite subtly that plug and play application of big language model will not yield a bounty which is being propagated. These would the peripheral benefits, though it would ease our present grind immensely. To make path breaking changes on these big AI enabled natural language processing multimodal models is to bring in a middle layer, between the platform and the API. This would be a large multimodal layer, which would mean huge amount of relevant tech work and its validation to create the final usable product. With this layer a GPT 3 or a DALL-E 2 would become ideally suitable for a sector or speciality.
You can thus have a model for Medicine. These customised models so to say in simple language, have the potential to completely change the landscape of a specific sector. Think of a company challenging the dominance of Google in search or Google reinventing itself. The new trillion dollar companies are likely to come from this fold. What would be the moonshots? Sam is of the view that moonshots would depend on the huge algorithmic progress that is in the offing. Multimodal models would become the order of the day. Its application would be a huge victory of tech revolution barely leaving any sector outside the throes. Who will take the moonshot? People are conversant with the this and broadly know its nature of growth. How do you commercially exploit the exponential decrease in the marginal cost of intelligence and also the marginal cost of energy is a trillion dollar question?
PREPARATIONS TO TAP AI FOR THE NEXT ERA HAS TO START WITH REGARD IN RIGHT EARNEST NOW.