Why is it important: Many believe that 2023 will be the year AI goes mainstream thanks to significant investment in any company or product that has “AI” or “machine learning” in its name. Microsoft’s renewed partnership with OpenAI does not confirm this prediction. However, it shows that the Redmond giant is moving from its failed attempts at mixed reality to big dreams for the future of AI-powered apps and services.
Microsoft has announced a “multi-billion dollar” investment in OpenAI, the artificial intelligence firm behind the hugely popular ChatGPT service and other projects like DALL-E and GPT-3. This discovery comes right after massive layoffs that have affected almost every team that previously worked on Microsoft’s metaverse and mixed reality projects.
The two companies have been secretly collaborating for years, and the latest moves suggest Redmond has a deep interest in OpenAI technology as a way to improve its software and cloud services ecosystem. The partnership began in 2016 but came to prominence in 2019 when Microsoft doubled down on a $1 billion investment made by OpenAI’s founders and other investors.
In subsequent years, OpenAI received approximately $2 billion and built its infrastructure on top of Microsoft Azure. Training and testing AI models require significant processing power, which is why Microsoft even developed a dedicated supercomputer to spearhead OpenAI efforts.
Both companies did not provide details about their new partnership goals. However, Microsoft says we can expect “new categories of digital experience” for consumers and businesses coming soon. Bloomberg notes that investments estimated 10 billion dollars over the next few years.
The rumor mill recently spilled something about Microsoft infusion its Bing search engine and its entire suite of Microsoft 365 apps powered by GPT-4, a yet-to-be-released artificial intelligence model that OpenAI is due to launch later this year. Given the backlash from companies like Google, the chatbot as the spiritual successor to the infamous clippy helper from the 90s does not sound so far-fetched.
There is a lot of hype around OpenAI technology in general and ChatGPT in particular. People who have used this tool find it capable of providing believable answers to a wide variety of textual cues. You can ask him to compose poetry, answer scientific questions, or even write code for an app or service. In other words, it can mimic how real people write and speak in a way that captures the imagination of people all over the world.
The answers given by ChatGPT are far from perfect, but some people worried the tool can quickly evolve to replace some jobs and create problems such as students using it to cheat on final exams.
Others, such as Yann LeCun, Meta’s chief scientist in artificial intelligence, are not so impressed with the current capabilities of the tool. During a recent virtual press event, LeCun said that “in terms of core methods, ChatGPT isn’t particularly innovative.”
While the scientist thinks it’s “well put together” from an engineering standpoint, he emphasizes that several organizations have developed various technologies that make it work for years.
In other words, it’s not the technology itself that’s impressive, but rather the scale of the data that was used to train GPT-3.5, the model that underpins ChatGPT in its current form. Asked why the likes of Meta and Google haven’t come up with similar tools yet, LeCun explained that both companies “have a lot to lose if they release systems that make things.”
Many machine learning experts agree with LeCun. Everyone agrees that generative AI tools such as ChatGPT have great potential for enhancing creative work, but there are many obstacles in the way of this goal. Examples include:
- Legal issues related to using copyrighted works to train AI models.
- Possibility of use in cybercrime.
- Relatively high chance that an AI model will give wrong and biased answers or unusable results.
Interestingly, even OpenAI CEO Sam Altman believes enthusiasm for his company’s technology should be reduced. Enthusiasts are already wild speculate about the long-awaited successor to the GPT-3, but Altman says they “hope to be disappointed, and they will be.”