Re: The complaint seems confused
"The tool is being criticised for being rubbish at accurately identifying gender. This may well be true but does it predict better than random?"
Indeed, there are a lot of complaints about it being wrong in specific cases, but there doesn't seem to be any evidence it doesn't actually work. The majority of professors are male. That may well be a problem with inequality in society, but that's not relevant to a system trying to make inferences based on the actual current situation. It says professors are male because most of the time that is likely to be the case. It's nonsensical to complain about a system reflecting reality because you'd like reality to be different.
Other examples may be more indicative of issues with the system, but it would still be nice to see some actual evidence that it's wrong rather than just the usual internet outrage. OK, it says someone with "woman" in their alias is 96% likely to be a man. The article claims that isn't true. It sounds like it shouldn't be true, but can you actually support that claim? Just because we don't expect, or don't like, some answers, that doesn't mean the system producing them must be wrong.
In any case, the most relevant part of the article would appear to be this - "Several companies have already been publicly providing similar technology for the last six years". Complain all you like about one company making it's tool free for the public to use, but there's no point worrying about how a trans person might react to seeing advertising aimed at women, because advertisers are certainly already using these tools. If you really think it's a problem, you need to address that, not just complain when you're correctly told most professors are men.