Image produced using MidJourney
We are seeing countries starting to be protective about AI developed in their country or hostile toward AI from other countries. Italy banned ChatGPT, although now reversed. China has a pretty clear lean toward locally-developed AI. India is thinking it through from a policy perspective. Here’s a list of countries that have banned ChatGPT.
IMO, India should not be left with no foundation models/LLMs initiatives developed in the country. There may be geo-strategic reasons for this. Much like having most of the world’s chip manufacturing in Taiwan is a risk now for the world. There is also some logic in having LLMs that do not show bias or leanings or concentrations toward one country or the other, perhaps based on the datasets they are trained on.
Having said this, let’s also acknowledge that the world of LLMs is very diverse and competitive and changing day-by-day. There is no “one LLM to rule them all.” There is an open debate on many aspects of this…
- American models versus European models versus Chinese models versus Indian models, subject to local rules around data, privacy etc?
- 26B parameters and hosted centrally or 7B parameters running on your smartphone? Sam Altman’s view on this.
- Open source versus closed source foundation models?
- Publicly-available models versus proprietary models running on private data?
- General purpose models or use-case specific models?
- Free models versus paid models?
- Models that create models?
We are actively speaking with founders pursuing opportunities in this area, including wrt regional languages and data localization rules, so please come chat with us.
Authors
