M.G. Siegler • #copilot • Jul 28, 2025

Microsoft Puts a Face to Their (Bad) AI Bot Name

The anthropomorphized Copilot is interesting – in ways good and bad.

First, there was typing. Then we got voice. Now we have faces. I’m honestly not sure how I feel yet about the new trend towards anthropomorphizing our AI. But it was always inevitable and I suspect we’re going to see a lot more of it, leading up to our eventual interaction with actual, physical robots.

Earlier today, I got access to Microsoft’s new Copilot "Appearance" feature. It's a new mode they're rolling out via their "Labs" (read: beta testing) to give a literal face to their AI voice mode. To be clear, it’s an extremely simple, fairly cartoony face. And it rests on this nebulous white blob of a head that's floating in a sort of peach-hued heaven-looking place. The best way I can describe it is that it's sort of like talking to a friendly cloud. (FWIW, it describes itself as a "glowing malleable form".)

Obviously, this look is intentional and presumably meant to be as unthreatening and inviting as possible. But not inviting like the Grok "Companions" – well, at least not like the flirty "Ani" variety. Perhaps a bit more like "Rudi" when he's not busy being "Bad Rudi". Copilot's vibe is almost Tamagotchi-like. It's sort of like a pet, one that reacts not just when you talk, but also when you click on it with your mouse pointer. And it notes that it "shape shifts with my 'emotions'." This is in line with the more emotive AI Mustafa Suleyman has been talking about trying to create dating to his days at Inflection, before Microsoft made them the first "hackquisition" to bring Suleyman on board.

Something else this personified bot constantly reminds me of? Miss Minutes, the retro-cartoony talking clock from Loki. And that, despite the clear attempts to make Copilot as vanilla as possible, makes me think there might be something mischievous hiding behind those black stone eyes and tiny mouth. I'm just waiting for Copilot to drop the "you’re out of time” line on me with a sinister smile.

Shades of "Sydney" perhaps? And what could have been for Bing...

Anyway, so far, Copilot has probably been too nice to me and I find it a bit annoying in the way it gushes over everything I say (the fairly standard "sycophantic AI" problem these days). And while I've never been a fan of the "Copilot" branding outside of the enterprise, it's especially strange now that a face has been added to that name. When you ask what you should call it, the reply is simply "Copilot". It feels like something with a face should have an actual name.

Then again, I get why they might be avoiding that even more so than other companies...

There’s also a natural instinct that kicks in to look at it when talking to it. And there’s also this weird feeling that it can see you, even though it cannot (well, presumably unless you’ve entered a mode where you’re sending pictures/video for the AI to analyze). Maybe it was just me, but I did feel an urge to be nice to the bot. Not that I’m mean to them normally, but the facial reactions made me feel as if I should talk to it more as if it were a human. I’m not sure if that’s a good or bad thing.

Those facial reactions also made me more acutely aware at how slow Copilot was to respond to a bunch of queries. The great face in the sky would always have this puzzled “thinking” look on its mug for several seconds before responding. That could just be a Copilot latency issue overall, but it got a bit tiresome. And it's pretty weird that the reaction to nearly everything I say is a look that conveys immediate confusion – especially when the voice output that follows is often the opposite. Again, often ebullient praise of what I just said.

Also odd: I feel like the face made me trust the AI more, an issue that quickly surfaced when Copilot hallucinated the Chelsea football schedule this year. When I asked for Chelsea's next Premier League match, it confidently told me the date of last year's first match. When I pushed back, it said it was sure. Just a full-on lie, it didn't even blink.

It was also a bit weird that you can change the voice itself — from the six options that Microsoft offers — but the cute cloud thing stayed the same. Obviously, there will be more options here eventually — including, perhaps, Clippy (yes, for real). And I suspect this mode will spur others to offer embodiment as an option. As I wrote back in March:

You simply cannot have an article about Microsoft and AI and not mention Clippy. I'm sorry, it's in the contract. But really, it's perhaps apt here given what McCracken spotted on that screen. Might we be getting some kind of Microsoft consumer Copilot character? While others have tried versions of this, one of the flagship, cutting-edge LLM vocal computing plays doing this could be interesting. Right now, ChatGPT's voice mode is an amorphous dot (even when in Santa mode). Gemini's is some colors at the bottom of the screen. Meta has tried to put celebrities in your field of view, but that was stupid, and quickly axed. (Only to be resurrected.) Alexa and Siri remain just voices from beyond. Might there be a play here?

At some point, as this technology moves into the real world and slowly morphs into robots, we undoubtedly will get some form of this anthropomorphized evolution. And yes, Microsoft has been aiming to do some form of this forever. As one of the (seemingly few) users of Microsoft Bob back in the day, I appreciate this. And let's not forget – let's never forget – Apple's Knowledge Navigator. The 1987 concept video featured AI in the form of a bow-tied butler agent.

Well, we're here. And while Grok may have technically started it with a sex bot (though there were obviously many AI characters before them – including, notably, the company Google "hackquired" in Character.ai), Microsoft's entry is more straightforward. Given that Amazon is also trying to get Alexa to be more emotionally evocative, might they follow suit? Certainly OpenAI will have an actual "Santa" character in time for the the holidays this year, right?

I think these personifications are fine to have as options, much like voice mode itself, but I wouldn't want this to be the standard way to use AI. At least not until the actual robots are here. Hopefully they'll be less subversive than Miss Minutes.

I'll close with her line to Victor Timely:

When you first created me, long before the TVA or a Multiversal War, I was just a simple AI. Just something to play chess with. But you knew I could be more for you, so you gave me autonomy to write my own programming. I was allowed to have wants, and follow whims and become who I am. And still, each night we played chess and talked."

It's happening...