A few days ago, reports surfaced online that OpenAI, the makers of ChatGPT may declare bankruptcy as early as the end of 2024.
However, now it seems that OpenAI will generate some pretty decent revenue after all and may not go bankrupt any time soon.
Backed by Microsoft, who invested over $10 billion in the company, OpenAI had projected that it would only manage to make just over $200 million this year. Last year, the company managed to generate only $28 million in its entire year of operations.
Changing tides
As per a report by The Insider, Sam Altman’s OpenAI is currently set to generate over $80 million in just a month and will continue to do so at least throughout the year.
The change in OpenAI’s fortune stems from the fact that it has broadened its approach and is now selling specialised licenses of its GPT-4 LLM, along with other AI software products. Most notably it is selling a lot of plug-ins as well as access to its API to companies who want to create their own AI chatbots, based on OpenAI’s GPT-4 LLM.
Core challenges remain the same
Although OpenAI is generating a healthy revenue, now, its core challenge remains the same. It is still burning through a lot of cash, just to keep the public version of ChatGPT, its website running. Reports have suggested that OpenAI burns about $700,000 dollars a day, just to keep the public-facing chatbot running every day.
Moreover, it has some tremendous costs for other, more intensive AI generators, like Dall-E2 and some of its other AI solutions
The problem of scaling an AI business
Scaling any business is a tremendously difficult proposition to be faced with. For Sam Altman and for OpenAI, the task is even more challenging. The more users that ChatGPT has, the more computational power it will need, which means investing in more data centres, more supercomputers, and more GPUs.
Plus, ChatGPT is facing some serious challenges from Meta and Google, and their AI solutions. Google and Meta, both, have released solutions that not only work faster than ChatGPT, but their enterprise solutions are also more accurate. Both Meta and Google are also offering their AI services coupled with a bunch of services that already have a ton of paid, enterprise users
Then there is the issue of drifting. ChatGPT is miserably bad at maths and is only getting worse, especially in complex mathematics, as per studies conducted by Stanford and various other universities. Large language models like GPT-4 have the tendency to get worse in certain areas when developers try to improve one specific area. No one knows why this happens, nor can anyone predict which in which areas will an LLM deteriorate.
The cost of all of this adds up, pretty quickly. While OpenAI may be generating a pretty solid revenue as of now, it certainly is miles away from sustainable profitability.
0 Comments: