Already there are many writings online discussing the potential of Gen AI in productivity, creativity and new discoveries. Massive investment thru VC to Gen AI startups and large corporations ramping up spending on this technology in the hope of automate their workflows / outputs.
I think this article "Who Profits the Most from Generative AI?" by MIT SMR, is a wonderful piece to understand the Gen AI Industry value chain. ie How every pieces stack-up to deliver an LLM application to companies & users.
1️⃣ Computing Infrastructures, where leading by NVIDIA, AMD, TSMC, Intel, etc
2️⃣ Data, technically almost the entire public internet, ebooks.
3️⃣ Foundation Model, leading by OpenAI, Anthropic, Google, Meta , where size-matters and founding capabilities.
4️⃣ RAGs and Fine-tuned models, which represent the service that aims to use an LLM for a specific purpose, to perform a context-specific tasks, with domain-specific data.
5️⃣ LLM Applications, like GitHub Copilot (coding), Sudowrite (creative writing), Alltius (customer support), etc... task-oriented offering.
The authors made a few "predictions"....
▶ The reasons that most cloud startup failed or were bought by rivals apply to generative AI as well --> Because continuous investment (competing resources & supply), Model size dictate LLM performance and economic-of-scale.
▶ Organisations that have access to lots of high-quality data in specific domains might have an advantage in creating specialised models. --> Strong Incumbents advantages, could be find in financial institutions, insurance, media and health-care, etc. e.g. Bloomberg
▶ Concerns around training models on copyrighted content are likely to result in a greater advantage for established players in generative AI. --> resources to battle copyright claims in court .... again $$$ matters.
For fellow transformaton leaders, keep an eye on 4️⃣ , 5️⃣ , where industry-specific & task-specific Gen AI may be an answer.
Also, if your company already have unique data and domain-knowledge, that enable and govern a fune-tuning or RAG, that may ultimately create an advantage. (but beware of leak of proprietary knowledge!!!).
There are other types of AI, may already promising in performance and less running cost & less energy consumption than Gen AI !! Be considerate on the choice. Cheers. 😉
#AI #GenAI
https://lnkd.in/e9qv46AB
Serial Entrepreneur, Advisor , Mentor , Independent Director with a passion to help startups
3moAnand Kannappan & Team Patronus AI - Congrats and am sure many more awards and recognition will follow 😎😊