Anthropic, OpenAI, and NVDIA
Parag Jagtap
Chief Technology Officer @Profount (Data Analytics) | Software Entrepreneurship | Transformational Leadership | Author | Board Member
Today I will be sharing two important developments and one rumor.
1) Anthropic Releases Beta Version of AI driven RPA
Anthropic is an OpenAI's rival founded by ex-Google and ex-OpenAI employees and funded by Amazon and Google. On October 22nd, Anthropic released a test RPA feature as part of Claude 3.5, latest version of its LLM model. Developers can access and use Claude 3.5 using APIs and build AI driven RPA (robotic process automation) applications.
This may not be an earth shattering news; however, it is an important one. RPA is now an important part of industrial automation. Using RPA, software companies have replaced hundreds of thousands of white collar computer related jobs that are repetitive; jobs such as order processing, invoicing, billing, payments etc. Current RPA is kind of rigid because it can only complete tasks that are repetitive and have fixed structure. The RPA software only knows where to click on a screen and in what sequence. It doesn't quite understand what's going on the screen. An AI driven RPA promises to make the RPA a lot more intelligent. An AI-RPA can intelligently achieve the work by processing information on the screen as opposed to current RPAs.
As of now, of course this feature is in beta/test mode and there will be errors. However, I expect AI driven RPA to become mainstay of RPA industry. This could eventually convert to a solid commercial success for Anthropic, in an otherwise year of AI disillusionment!
2) Honeymoon over - OpenAI and Microsoft
There were rumors about OpenAI and Microsoft not being quite on the same page. Apparently OpenAI is going to be $5 billion dollars in loss this year. It is asking Microsoft to open its wallets and fund OpenAI further. Microsoft has already invested $13 billion in OpenAI since 2019 and owns 49% of the commercial entity of OpenAI. It is probably not thrilled to continue to sink large sums of money into OpenAI.
Being an AI infrastructure provider is a costly business. There are only a handful entities in the world that are attempting to be one. OpenAI, Anthropic, xAI, Google, Microsoft, and Amazon are the major players. Companies developing LLMs and investing in general intelligence are building large datacenters or renting such datacenters to mine huge quantities of data and build these LLMs to manufacture intelligence and enable AI agents. It requires costly NVDIA GPUs. Elon Musk famously has bought 100,000 NVDIA chips for his X data center. Each one of these GPUs cost tens of thousands of dollars. So that could be a $1-3 billion order of GPUs.
As I said, producing artificial intelligence is costly.
All these providers are hoping to sell intelligence to various industries around the world. OpenAI has monetized well so far. It is nearing a run rate of $4 billion revenue per year. Yet they expect to lose $5 billion in 2024. The cash burn is too high.
Since 2019, the dynamic between Microsoft and OpenAI has changed. LLMs are no longer as much a mystery today as they were in 2019. Cost of computing is a major barrier to entry in building larger and more capable LLMs. But that is not at all out of reach for Microsoft. It could very well develop its own LLM - which is exactly what it is doing.
It is possible that OpenAI is only now realizing that by allowing Microsoft - a potential future competitor - to own 49% of OpenAI, it may have become too dependent on Microsoft for its own good. From Microsoft's point of view, the investment in OpenAI is already producing returns because ChatGPT is hosted on Azure, the Microsoft cloud. Today, as LLMs are not a mystery, Microsoft can potentially develop and sell competing products. Having decades of experience revolutionizing the world with new technologies, Microsoft could be in a far better position to build and sell AI infrastructure than OpenAI.
I honestly can't figure how this will end. My sense is that OpenAI converting a cross industry not-for-profit to a commercial entity may not end well for either the commercial arm or the not-for-profit arm due to the inherent conflict between the two.
3) NVDIA held an AI summit in India
NVDIA held an AI summit in Mumbai, India, as India gets ready to benefit from AI revolution. NVDIA signed a number of deals with tech titans in India as well as industrial powerhouses like Reliance. Reliance Industries, is considering betting big on AI revolution and partnering with NVDIA to build AI infrastructure for India. Companies like Tech Mahindra are partnering with NVDIA in order to build LLMs in local languages to help build local AI solutions in local languages. The Indian government too won't be left behind and hinted that discussions are going on to develop an AI chip for India in partnership with NVDIA.
It is true that India has become the back-office for the whole world and has a large base of software engineers trained in AI. A lot of those engineers could already be helping develop some of the most cutting edge technologies for American or European companies. This ready talent will be useful as India starts developing AI solutions local to India. As most Indians speak in native languages; solutions equipped with native languages and context will be required to serve them.
Just like in USA, Indians are also skeptical of their own government. Government involvement is a recipe for delays, mismanagement, and corruption. So it would be interesting to see how the AI chip project unfolds. It is critical for a large country such as India to have its own chips as this is a matter of national and economic security and sovereignty. It would be even better if one of the Indian businesses, rather than government, partnered with NVDIA to bring such chips to market. Regardless of who does it, it is important for India to have its own AI chip. On a separate note, my view is that, India also should guard its own data and put control on foreign social media like Facebook, WhatsApp, X, and tikTok. But that's a slightly different topic.
In my humble opinion, Indian government can put AI for great use for transforming itself. There are two key areas that I can think of. I am sure there will be many more.
1) Using AI, India can significantly reduce fraud and corruption within government and in economy. AI is famously good at detecting fraud and anomalies.
2) Using AI, India can drastically reduce backlog of legal cases. India's judicial system is archaic, manual, and colonial. India has a case backlog of almost 51 million cases, out of which about 200K cases are multiple decades old. Part of the reason why the judicial backlog is so high is becase there is always a shortage of judges. AI could come in and help deliver justice quickly, efficiently and by the book!
Let's hope India doesn't remain just a consumer of AI. Let's hope it becomes a producer of AI and guardian of its own data.
That's all for today. Your comments and feedback is welcome!
References-