The Internet's largest chat app is going all in on AI, and at a virtual "AI day" event hosted by Discord for the press, the company announced three experimental AI tools that will soon be tested on select servers. The most important of these is a bot that uses OpenAI's versatile chat interface, ChatGPT (opens in new tab), which Discord plans to integrate into all of its servers at some point.

Discord's ChatGPT bot will be named Clyde and will be able to ask him anything by "@ing" him in the server, just like a real person. However, instead of politely ignoring you or waiting 15 minutes for a response like a normal human, Clyde will respond back to you within seconds.

In a Discord demo I attended, a user asked Clyde for the current time and weather in Tokyo. Simple enough, they asked Clyde for "a good superhero movie gif that expresses the joy of watching a movie with friends." When looking for a gif that expresses the joy of watching a movie with friends, it would take five minutes to pick one that would make them laugh or feel clever for finding it. Clyde experienced no such restraint and immediately offered a gif of Rizzoli and Ailes lying in bed from the 2010 TNT network hit crime drama "Rizzoli & Ailes," with the caption "WATCHING MOVIES." Clyde added, "It's always super fun to watch movies with friends," and I couldn't agree more.

A year ago, I might have thought Clyde's suggestion was kind of cute, but the "magic" of AI chatbots is eroding at record speed; talk to ChatGPT long enough and the novelty wears off and you find yourself interacting with a somewhat conversational search engine. (Open in new tab), but there's no personality to connect with it, and no need to have a chatbot do basic Googling for you. The same goes for Clyde.

If you don't like Clyde interrupting your conversations, you can always turn him off in the server settings. Clyde will be available as a "public experiment" on some servers as early as next week.

Discord already has an Automod tool that automatically flags messages that may be in violation of the rules, but according to the company, this tool will soon be upgraded to AI and will be able to interpret the context of a message in addition to following the letter of the law

The idea is to make it easier to understand the context of the message.

The idea is to reduce false positives and catch offending messages that try to circumvent the rules; Discord demonstrated a cat-only server where one user tries to circumvent the rules by talking about fish, but spells "fi$h" to avoid the Automod flag Automod catches this not-so-subtle spelling; according to Discord, Automod can also tell the difference between, for example, a user calling for "death to all hamsters" (against the rules) and a user sharing the main cause of death of a hamster (allowed ) can also tell the difference, according to the company.

Interestingly, Discord also states that Automod's AI has been enhanced to enforce even unstated rules; Discord claims that Automod reads the rules written in the server's official text channel "#rules" and claims to be able to abide by them.

This particular claim was not clearly demonstrated in the video shown to me. Such a claim sounds too good to be true, or at least too ambitious to work well in practice. Having moderated a chat room with a large group of people before, I can see the value of an AI assistant that can read between the lines, but it seems super risky to give an AI the keys to moderate outside the strict parameters set for it. What if your rules section included a fake rule like "absolutely no Coca-Cola talk" that a human could understand as a joke, but an Automod would take it seriously; the flagged sanctions shown in the Discord demo would at least be justified by a human who needed to be confirmed.

While Clyde and Automod are the furthest along in the AI affair for Discord, it also showed a "sneak peak" of several other tools in the works, including several Discord-assisted projects from outside companies:

The event also featured 2.9 million demos of Midjourney, an AI art-generating bot active on more than 2.9 million Discord servers, and Discord spent much of its presentation praising Midjourney for fueling the art revolution and giving people new ways to be creative. I used to feel similarly about the potential of AI art, but now generating jpegs with keywords sounds like the opposite of creative.

I think we are all starting to develop a radar that the artwork we scroll past is probably AI generated, the Midjourney bot makes a quick impression, like quickly generating a photo of a forest to set the mood for a D&D session I think it's a useful tool, but otherwise it's hard to care about AI art as much as Windows 11's default wallpaper.

Meanwhile, a feature that probably makes more sense for the average Discord user just rolled out yesterday: official PS5 voice chat support (opens in new tab).