Microsoft in January made its third round of investment—$10 billion over a multiyear period, according to Semafor—in OpenAI, a company that specializes in artificial intelligence, specifically large language models.
OpenAI’s ChatGPT has captured the attention of CEOs, software engineers, and the general public and was the quickest technology service to gain its first 1 million users, which took just five days, according to CEO Sam Altman.
Although Microsoft has an early lead in AI, we expect other software vendors to replicate its moves: striking partnerships, building their own AI models, and/or using application programming interfaces to incorporate AI into their own products.
Though still in its infancy, the introduction of advanced generative AI has far-reaching effects. And has raised concerns among consumers and governments around whether humans can control such technology over time.
The world is ending. AI must be stopped.
A review of some of the headlines in the popular press suggests that the world is ending. Governments have passed or are contemplating passing laws that prevent or curtail the capabilities that OpenAI brings to bear.
We acknowledge that there are several potential risks that may come from AI if left unchecked, including potential job losses due to automation, social surveillance and manipulation, misuse, violation of intellectual property rights, biases and inaccuracies in training data resulting in biases and inaccuracies in output, and a variety of worst-case-disaster scenarios—unlikely as they may be—such as autonomous weapons systems being commandeered by AI systems or the breakdown of the financial system.
But we think these fears miss the mark, at least with OpenAI’s products to date. In our view, the way to mitigate these risks will include domestic and international regulations as well internal corporate standards, controls, and disclosures. We think trial and error in model development and usage will be an important factor as users become more comfortable working with these new advancements. AI is at the forefront of technology and is therefore likely to require creative guardrails, in our view, but we doubt there will be a mechanism or will to do much to slow proliferation.
Ultimately, ChatGPT is just another software tool. It helps users perform a task more quickly than they otherwise would be able to without it. We have not seen generative AI solutions broadly introduced as features or adopted, so we think fear is premature.
Perhaps the most cited concern is that generative AI may replace knowledge worker positions. But again, these large language models and their interfaces are just tools. They help write code faster, summarize meetings, and streamline customer service by handling basic requests or problems. They make professional jobs easier by improving efficiency. Generative AI helps by handling more mundane tasks so knowledge workers can focus on the value-added parts of their jobs. It helps employers improve worker productivity.
Further, even if some existing jobs are replaced, AI itself will likely create jobs. AI models are software-based and must be written, trained, and maintained. Web developers, for example, did not exist before the internet. Content creation on the internet certainly has pressured legacy content creation for magazines and newspapers. We will reconsider our position on this when we start seeing entire professions suddenly out of work.
On a related note, technology companies have lamented for years that finding capable software engineers is a real challenge. Given the specialty required for generative AI development, we suspect capable engineers will be even harder to find and will likely command premium salaries.
Everybody will have AI, so this is no big deal.
We think the stance that generative AI is no big deal is also off the mark. This technology is powerful, and we think it will be a useful tool for improving efficiency for knowledge workers. We can see new revenue streams for Microsoft and other software vendors, even if it’s only incremental revenue. We can see many tasks where generative AI will make certain professions easier and help streamline a variety of tasks. This is not inconsequential.
Conversely, the market reaction has been to send Microsoft’s shares up, along with those of many technology peers. Technology in general has performed reasonably well, and Microsoft just reported an excellent quarter that had little to do with AI. So maybe the OpenAI announcement is driving the shares up, but maybe not.
We have no idea what to make of AI just yet.
We think this is a reasonable camp to be in. We view the technology and competitive landscape as being in its very early days. The potential for change is enormous. A well-established market with a variety of competitors and products does not exist.
Estimates are that OpenAI generated $50 million in revenue in 2022. It’s a big leap to extrapolate that to “the world is ending,” but we would also caution investors that this is not “no big deal,” either.
We believe that advanced generative AI will soon be offered across virtually all software products. Some believe that AI will depress margins as companies have to invest to keep up. We do not think this will be the case; at the very least, we expect the margin pressures resulting from AI investments will be manageable. We think larger companies can develop AI as another research and development project within a portfolio of such programs. Further, many companies will simply be using OpenAI, or one of its competitors, through an application programming interface. In these cases, the software company would most likely be charging the end user for this in a premium pricing tier, thus maintaining its margin structure.
Software companies are racing to bring forth their own advanced generative AI features. Various consultancies, internet firms, and semiconductor companies are also well positioned to capitalize on the AI boom. Many companies are already touting their ChatGPT API integration, while others have introduced their own unique and powerful AI instances. We think the largest companies we follow have the most resources at their disposal and are thus better positioned to build, train, and maintain AI algorithms. And we think independent AI developers like OpenAI will license their technologies to other companies, thus making highly advanced AI models available to smaller companies as well.
Frankly, we are at the beginning of this movie rather than the end—or even the middle, where we might have a sense of the plot.