Investors should get used to extreme competition in the artificial intelligence (AI) arena. According to analysts at Bank of America, artificial intelligence will add $15.7 trillion to the global economy by the end of this decade from both increased revenue opportunities and lower costs. That means AI capabilities will be table stakes for any company hoping to succeed.
Two chipmakers leading the charge are Nvidia (NVDA -1.99%), which has first-mover advantage, and Advanced Micro Devices (AMD 1.14%), the scrappy challenger that more recently came out with its MI300 Instinct AI accelerators.
This week, both companies made announcements for their latest parlays in the ongoing quest for AI supremacy. But of the two, one company appears to have the clear advantage.
AMD buys open-source AI software firm Nodi.ai
On Tuesday, AMD announced it was acquiring open-source AI software start-up Nodi.ai for an undisclosed sum.
Nodi.ai was initially founded in 2013 as an AI hardware company, aiming to give gamers the ability to control game elements using only their fingers instead of a controller. However, the company has since pivoted to open-source AI software, which can also run on AMD's chips.
In the press release, Nodi.ai CEO Anush Elangovan said: "Our journey as a company has cemented our role as the primary maintainer and major contributor to some of the world's most important AI repositories. By joining forces with AMD, we will bring this expertise to a broader range of customers on a global scale."
And AMD Senior Vice President Vamsi Boppana added:
The acquisition of Nod.ai is expected to significantly enhance our ability to provide AI customers with open software that allows them to easily deploy highly performant AI models tuned for AMD hardware. The addition of the talented Nod.ai team accelerates our ability to advance open-source compiler technology and enable portable, high-performance AI solutions across the AMD product portfolio. Nod.ai's technologies are already widely deployed in the cloud, at the edge and across a broad range of end-point devices today.
The acquisition follows AMD's acquisition of French AI software start-up Mipsology back in August. Mipsology is another AI software stack specializing in photo and video recognition and geared for applications such as autonomous driving and smart manufacturing robotics.
AMD is not only aggressively investing in its own ROCm AI software stack with internal investment but also making these bolt-on acquisitions in the space to add new capabilities. According to TechCrunch, AMD now has 1,500 engineers in its recently formed AI group and will add 300 before the end of this year. It appears Nodi.ai and Mipsology will become part of this group.
But while this week's news could be seen as a positive for AMD, the fact that it's still having to cobble together capabilities via acquisitions indicates it's still well behind Nvidia and spending hard to catch up.
Image source: Getty Images.
Meanwhile, Nvidia continues to innovate at a furious pace
Nvidia started its own closed-source CUDA software stack all the way back in 2006 and continues to innovate in AI at a fast pace from a leadership position. An example came on Wednesday, when the company introduced a new AI inferencing capability called Nvidia NeMo SteerLM.
The new software tool allows developers to greatly streamline model building for inference. After training a large language model (LLM), an inference model dictates how an AI specifically reacts to a stimulus.
LLMs are massive and contain a ton of information. In order for developers to apply them to actual specific use cases, they need to give the LLMs "directions" when prompted. While more and more developers have done this, the process is time- and labor-consuming.
SteerLM greatly streamlines the process, allowing developers to tune an inference model with just a basic set of prompts. Those prompts are then run and produce a data set. And that dataset can then be used as a baseline input for an AI model that can be easily adapted on the fly.
If that sounds complicated, Nvidia gave some examples of how this may be applied. One example is an enterprise customer service chatbot that can alter its parameters on the fly, given a customer's age, geography, and even softer characteristics such as changing attitudes.
Another example is for a law firm or enterprise to output answers from an LLM in either formal legal text or a more accessible conversational style, depending on the audience. And game developers can have characters in games react organically to a player's actions or circumstances, rather than the traditional, mechanical pre-ordained list of actions or speeches pre-programmed by a developer.
Nvidia still seems far ahead
Most companies and cloud providers would love more competition in the AI space and will no doubt support AMD's efforts to catch up here. However, this week's news flow seems to show that AMD is still making foundational moves to boost its capabilities, while Nvidia seems to be moving forward with rapid organic innovation from its in-house teams while making artificial intelligence closer and closer to human interaction.
While AMD will no doubt get some piece of the AI pie, it doesn't seem like it will catch up to Nvidia anytime soon.
