Rs 10K Cr Infusion For AI Positive, Ongoing Support Now Crucial: Ayush Singhal, CodeMate

Ahead of BW Businessworld’s Gen AI Summit on 27th March, CodeMate Founder Ayush Singhal discusses the company’s unique strengths, comparison between Claude 3 and GPT-4 and India’s Rs 10,300 crore financial infusion for national AI mission.

CodeMate was founded in 2022, just before OpenAI’s ChatGPT came to the fore and disrupted the tech industry and beyond. You address the area of coding, how are you differentiating yourself in a post-ChatGPT and post-Copilot world?

We believe ChatGPT and Copilot’s entry into the market facilitated awareness among companies regarding the potential of Generative AI (GenAI) solutions in software development. This heightened awareness actually translated into increased inbound leads from companies seeking GenAI implementations, which helped us as companies wanted to implement Generative AI (GenAI) solutions.

Unlike Copilot and ChatGPT, we position CodeMate as the solution for post-initial code generation tasks. While aforementioned tools excel at generating initial code snippets swiftly, our focus lies in facilitating code modification, error fixing and contextual adaptation across entire organisational codebases. Our differentiation extends beyond functionality to include a crucial aspect: self-hosting. Unlike tech giants, who rely on centralised hosting, we empower clients with full control over their models and data through self-hosting options. This focus to client autonomy resonates with companies increasingly concerned about data privacy and security, a stance that sets us apart in the AI coding tool arena.

Have the AI models boom helped CodeMate? 

You can actually connect your own model with CodeMate. We are not an LLM-making company. What we work on are ‘searching algorithms’ and how effectively one can give the context to the model. We believe LLMs are just a like a book-smart-programmer, which knows the context of many books, but isn’t aware of the context. We work heavily on the giving the right set of context to the models. If people bring their own models, we have an option of plug-and-play solution. We simply link your model or GPT or any proprietary model to CodeMate and CodeMate helps you solve coding problems.

Does Claude 3 seem better to you than GPT-4? There have been reports of executives preferring Claude 3 over GPT-4…

Absolutely. When it comes to contextual searching, the advent of GPT-4 brought its own challenges, including hallucinations. However, with Claude 3, there’s been a notable improvement in inference speed and response quality. GPT-4 suffered from sluggish performance despite decent quality responses, leading developers to prefer GPT-3.5 for its faster speed. Claude 3 addresses this issue by delivering rapid responses without compromising on quality. In fact, in some cases, its responses surpass the quality of GPT-4. Claude 3 resolves the dilemma developers faced, offering both speed and quality in AI-generated responses.

A Tel Aviv University professor told us last week that he hopes the Rs 10,300 crore investment outlay from Indian government should be dedicated to research and startups, but more funding will be needed to rival US, China or even companies such as OpenAI and Google. Your thoughts?

Indeed, the government’s investment of Rs 10,300 crore is a significant boost for small startups and will foster dedicated AI ventures and innovation nationwide. However, to compete with tech giants like OpenAI, Perplexity and Google, substantial long-term investment in the billions is necessary. Building foundation models entails exorbitant GPU costs, beyond what Rs 10,000 crore can cover. While this investment is a positive start, ongoing support from industries and governments is crucial. As startups pioneer innovative ideas, continuous investment from the government is essential to sustain momentum and drive long-term growth in the AI sector.