Not to be outdone by the rising (well, sometimes falling) star of Google Bard, OpenAI has announced that popular language learning model ChatGPT now has official plugin support for developers.
Under OpenAI’s ‘iterative deployment philosophy’, the AI lab is releasing initial plugin access to a limited number of partner companies in order to study how the plugins are used, and how effective they are, before committing to a wider-scale rollout (developers can currently subscribe to a waitlist for access).
We have a list of the companies that have been granted access, and have already created plugins: Expedia, FiscalNote, Instacart, KAYAK, Klarna, Milo, OpenTable, Shopify, Slack, Speak, Wolfram, and Zapier.
That’s… a really weird list. I’m not the only one who thinks that, right? Some of them make sense; Wolfram and FiscalNote are both organizations that have history working with AI technology, while ChatGPT integration with Slack was something we knew about already. Speak isn’t entirely surprising either; we’ve already seen GPT-4 working as a virtual tutor in rival language app Duolingo.
Other participants in the plugin program feel a little harder to justify. OpenAI says that it’s ‘excited to build a community shaping the future of the human–AI interaction paradigm’, but I’m really not convinced these are use cases where human-AI interaction is actually needed. I can already book a holiday quite easily using the KAYAK app – I don’t need a chatbot to handhold me through the process. Do you?
Opinion: ChatGPT has its uses, but it doesn’t need to be everywhere
I feel quite dubious about several of the plugin-touting companies on this list – I don’t think shoehorning AI into restaurant-booking app OpenTable was necessary at all – but there’s one that really worries me: Klarna.
Now, I’m not here just to rag on Klarna (or any buy-now-pay-later service), but let’s be honest here: these platforms are moderately helpful at best, and downright predatory at worst. I’m sure not everybody will share my view here, but I’ve always felt that companies such as Klarna, Afterpay, and Zip exist solely to prey on people living paycheck to paycheck – getting into contractual debt over smaller expenses like ordering takeout or buying some groceries is basically never a good idea, if you ask me.
The point I’m trying to make is that some of the people who suffer the worst at the hands of pay-later companies are exactly the same people who will be more vulnerable to the encroaching effects of AIs in our society. I’ve written before about how AI could ruin a whole generation of future kids if we’re not careful, and slapping ChatGPT into Klarna is exactly the sort of dodgy nonsense I was talking about.
AI is still a largely unregulated space; there’s nothing to stop ChatGPT from suggesting that you absolutely should take out a loan to get that burrito delivered. Of course, ChatGPT doesn’t actually know what a burrito tastes like, but it sure can amalgamate a bunch of food critics’ words to convince you that it does.
Honestly, this sort of thing worries me. Not on a personal level, because I’m not about to let a chatbot convince me to get into debt over a shopping cart of food, but on a moral level, because some people will be swayed by it. There’s also the entirely different threat of companies like Klarna booting out human customer service in place of AI assistants; not only could that cost real people their jobs, but debt-stricken, desperate individuals will be better served by an actual person than by an unfeeling chatbot.
Look, AI isn’t all bad, and ChatGPT isn’t evil. And sure, OpenAI needs to make money; running a large, complex machine-learning program like ChatGPT requires a lot of hardware and a lot of energy. But we need to collectively take a serious look at how far we’re willing to go when it comes to letting it into every facet of our lives. AI isn’t the big problem – but how people use it definitely is.