AI Competition Reaches Diminishing Returns as GPT-4o is launched
From Abhivardhan, our Chairperson
This is a post authored by Mr Abhivardhan, our Chairperson.
So, I came across a text evaluation chart provided by Gary Marcus on X (formerly Twitter) on GPT-4o.
As he rightly pointed out - GPT-4o isn't that much better than GPT-4 Turbo and GPT-4 (Initial Release of March 14, 2023), and is slightly better than Claude 3 Opus, which I enjoy using in my day-to-day tasks.
As rightly pointed out by Gary Marcus on X/Twitter, OpenAI and their competitors have reached a phase of diminishing returns.
Also, as wrongly pointed out by some pointless and popular influencers on this platform - we are not living in some AI-driven internet era. Recommendation algorithms in SEO, social media and adtech for data scraping has matured since the 2010s already. Yet, people hype because it suits their "likes".
Let's understand how does that affect India's AI ecosystem.
1️⃣ We have to build large models and instead of obsessing around use case apps etc, which can then be remodelled and copied by someone better than us - it is time we begin with technical benchmarking and industrial standardisation of AI. Every app built on a #LLM is a feature, which can easily be copied. Most don't get it.
2️⃣ We at Indian Society of Artificial Intelligence and Law clearly believe that such form of standardisation of AI cannot be achieved with government efforts in #India since it would be a fait accompli situation for a government to be judge, jury and executioner at the same time. At best, some intermediation of certifying standards can be done by sector-specific regulators like RBI, ACI, and others. Plus, we do not have good jurisprudence on AI in India yet.
3️⃣ Governments in India and the Global South will have to face with the shortage of GPUs and the cost issues associated. There are economic rights issues associated with Indian businesses, especially startups and MSMEs. We cannot build an #AI use case capital in #Bharat unless we are clear we would support our ventures with conviction.
4️⃣ We have a lack of imagination in #Delhi because adding all major institutions, we have less than 500 GPUs (v100 and above) across the IISc and IITs. That is alarming. There is crunch in resources beyond big tech companies / MNCs, and that must be addressed by the Government of India if we are serious.
Meanwhile, do check out Technology Law Fundamentals, a training programme by Indic Pacific Legal Research to help tech professionals understand the basics of technology law: https://www.indicpacific.com/challenge-page/techlaw101
[New Course] Technology Law Fundamentals by Indic Pacific Legal Research
There's been some fresh debate on what is #legaltech and what is #techlaw. Let's unpack.Indian Society of Artificial Intelligence and Law is a reader-supported publication. To receive new posts and support our work, consider becoming a free or paid subscriber.