The world of Artificial Intelligence (AI) is in a state of flux as smaller companies are quickly catching up with big tech giants such as Google and OpenAI. This development could pose a significant challenge to the latter by advancing rapidly and outpacing these bigger corporations. This is particularly notable because these projects are executed with limited resources, often on penny-a-minute servers and laptops.
In the AI industry, it’s not new to see this type of disruption happen regularly, but a recently leaked document purported to be from Google put the matter in perspective. The memo suggested that Google, and indeed OpenAI, lacked a “moat.” Essentially, this means that their edge was becoming slimmer day by day as smaller companies with open-source projects began to catch up.
For example, when LLaMA, a foundation language model, leaked in March, people tinkering on their laptops quickly improved the rough draft. These enthusiasts added core features from human feedback, like multiple modalities, reinforcement learning, and instruction tuning. The code was even accessible to small companies, enabling them to collaborate and experiment more quickly than the deep-pocketed corporations.
It’s possible that the significant computation problem that initially posed an insurmountable obstacle – a moat – to challengers is becoming a relic of a different era of AI development. Sam Altman, an entrepreneur and former co-chairman of OpenAI, stated there would be diminishing returns when throwing parameters at the problem.
The prevailing business model used by companies like OpenAI is similar to the SaaS (Software as a Service) model. They offer carefully gated access to high-value software or service through an API. It’s a proven approach that makes sense when you’ve invested millions into developing a monolithic yet versatile product like a large language model.
However, customers have been raising questions about the necessity of employing the services of the most significant and universal AI model ever developed, mainly when their needs are limited to comparing the language of a contract with just a few hundred others. Furthermore, they argue that using such a powerful AI model for comparatively simple tasks might not be cost-effective and efficient.
In the AI world, a large language model was quickly run in truncated form on a Raspberry Pi. For companies like OpenAI, Microsoft, and Google, this effectively begs the question of their existence as it debunks the entire premise of their business. The companies chose and engineered a version of AI that fit their existing business model rather than the other way around.
Once upon a time, the computation involved in word processing had to be offloaded to a mainframe because terminals were just displays. However, as technology advances, we can now fit the entire application on a personal computer. This process has occurred many times as devices have exponentially increased their capacity for computation. In the AI world, this time has come much quicker than expected, and the big tech companies were not the ones doing the optimization – and may never be at this rate.
Nevertheless, big tech companies like Google have some advantages, as being like Walmart has its benefits. Businesses may prefer to avoid seeking a customized solution that could potentially perform a task 30% faster, opting instead for a reasonable price from their current vendor to maintain stability. Never underestimate the influence of inertia in the business world.
It’s worth noting that open-source projects run by smaller companies are not without their drawbacks. They are often incomplete and require users to work with them to improve their functionality. However, given the rate at which these projects are being improved, it’s only a matter of time before they become more effective than giants.
The rapid development of open-source projects is changing the landscape of AI development. OpenAI and Google are likely to continue to dominate the AI landscape for years to come, thanks to their vast resources and extensive experience in the field. However, the rise of open-source AI projects could spell trouble for these tech giants, as smaller startups and independent developers are able to innovate and iterate at a much faster pace. As AI continues to evolve and mature, it’s not confirmed to say who will come out on top. But one thing is sure: the AI revolution is just starting.