ChatGPT woke the world up to what was possible with AI, but its most enraptured fan is OpenAI. The chatbot has become OpenAI’s baby, with the company doting on this product to help it push its valuation from $260 million to $20 billion. However, it seems that this push for ChatGPT has alienated many of the company’s employees, as they are now leaving to find greener pastures. 

According to media reports, some key OpenAI employees have resigned from the company and signed along the dotted line with Google DeepMind. The exodus isn’t stopping there, as even more are slated to follow suit in the coming days. This move mirrors the initial brain drain that Google DeepMind saw last year, with OpenAI poaching employees from the tech giant to build ChatGPT. 

While this seems like a classic case of reverse brain drain, the departing employees have also shed some light on the shortcomings of Sam Altman as the CEO of the company. Not only does this echo some of the criticisms levied against Altman in the past, but it also sheds light on some of his regulatory moves, pulling back the curtains on OpenAI’s secret show. 

Mismanagement and mixed priorities 

In 2016, OpenAI had a grand total of six employees. Now, after aggressive hiring and growth, the company boasts over 600 employees. It looks like the company has not adapted its startup culture to this explosive growth, leading employees to leave the company. What’s more, with Microsoft’s $10 billion funds, OpenAI seems to be content chasing after their biggest cash cow, ChatGPT. While enough and more has been said about this chatbot, this repetition is likely getting to the people creating it too. Frustration behind planned further development of ChatGPT has resulted in employee dissatisfaction. 

This could be due to perceived stagnation amongst the employees as OpenAI’s pace of innovation was extremely fast before the launch of ChatGPT. A now-deleted article also showed that OpenAI’s roadmap for the next year only focused on growing ChatGPT by improving GPT-4. 

Interestingly, these statements came from CEO Sam Altman, who is also cited as another one of the main reasons that employees are leaving the company. Reportedly, employees stated that he only has a ‘superficial understanding on many topics and hardly cares about day-to-day business’. Moreover, they also said that Altman’s blanket statements on the dangers of AI are ’accessories to appease politicians’. This criticism has been levied against Altman in the past, especially when it comes to creating favourable regulation for OpenAI’s algorithms. Many state that he is jacking up the perceived disruptive impact of AI in a bid to push for a more conducive regulatory environment for big techs, leaving smaller companies behind. 

Amidst mismanagement, it seems that OpenAI is also losing its reputation in the research field. The company had an excellent track record of research which has slowed down greatly, going from a peak of 39 papers published in 2018 to a measly five in 2023. 

Engineers are a finicky breed, going after money, prestige and, most importantly, ideology. Google has been projected as the best place to work for decades now, and also presents a prime opportunity to jump ship for any engineer. Moreover, with Demis Hassabis, the CEO of DeepMind, announcing that its next algorithm will ‘eclipse ChatGPT’, engineers have the opportunity of a lifetime to be in the right place at the right time. 

DeepMind back in action?

In late 2022, as OpenAI wowed the world with ChatGPT, other companies, especially DeepMind, were seen as falling behind. However, this changed with Google’s FY22 Q3 earnings call, as it announced that they would integrate DeepMind into itself to supercharge its AI efforts. Looks like this restructuring has paid off as Hassabis has now announced a new model.

According to reports, this large language model called Gemini will combine the typical transformer architecture with the techniques used in AlphaGo. This will give the algorithm better problem-solving capabilities, along with a host of other improvements. Hassabis stated, “At a high level, you can think of Gemini as combining some of the strengths of AlphaGo-type systems with the amazing language capabilities of the large models. We also have some new innovations that are going to be pretty interesting.”

As the model uses the transformer architecture, first pioneered by Google, employees from OpenAI could not only add value to the new project, but might also delve deeper into OpenAI’s optimisation techniques. Combined with DeepMind’s existing expertise in deep reinforcement learning, Gemini might become the first LLM to actually blow GPT-4 out of the water. This could have been one of the biggest draws for OpenAI researchers to join Google, and the company might have even sweetened the deal. The company is currently hiring research scientists and engineers en-masse for its London office, which seems to be the move that triggered OpenAI’s exodus. 

Even as OpenAI is currently in the limelight, this brain drain could slowly dilute the value the company is bringing to the market. Many have also argued that OpenAI’s new-found funding is causing the company to lose sight of its original goal, something Altman himself has refuted in the past. However, when comparing Altman with Hassabis, an AI expert with multiple papers under his belt, it is clear who the scientists and researchers will prefer to work for.