When AI does it for you, is it lying to your clients?
Microsoft 365 Copilot was made available to a limited number of businesses in March and is now being offered to more businesses for an unknown amount. This takes Microsoft’s new business model to a new level. Instead of giving you the opportunity to beta test for free, you get the opportunity to pay for the privilege of testing your software before it’s ready for general release. If they’re suckers enough to buy into this plan, they can unleash the power of ChatGPT-4 on their colleagues and clients.
If you haven’t paid too much attention to the behind-the-scenes findings of LLMs such as ChatGPT, LLMs have become quite famous in some circles for their hallucinogenic abilities. It’s the official term given to AI applications that falsify data, invent non-existent quotes, provide contradictory answers, and generally lie to those who use them. LLMs, aka AIs, have no qualms about stealing poorly protected personal data or writing spectacularly insecure code. Microsoft 365 Copilot brings these capabilities to the corporate world. The corporate world has already taken full advantage of the aforementioned capabilities without the need for electronic assistance.
Features available to subscribers include a search capability called semantic indexing. This searches for keywords instead of searching for keywords, but adds irrelevant context to the results. Of course, this assumes that the algorithm won’t invent an answer if it can’t find a result that it considers to be of high enough quality. And by incorporating OpenAI’s DALL-E text-to-image generator into his PowerPoint, it gives you the opportunity to unknowingly steal copyrighted art and incorporate it into your slide deck.
Microsoft, of course, is slapping the warning that some of the data Copilot generates is inaccurate. It is unlikely that you will receive it.
