With all the hype surrounding ChatGPT, it’s no wonder other companies are vying for a piece of the AI-powered chatbot game. Companies are betting that we’re at a decisive moment in the artificial intelligence industry, where products that adopt and build upon the budding technology could have the potential to reshape technology as we know it — not to mention shake up the Big Tech hierarchy.
The stakes are high, and technology’s biggest players don’t want to be left behind as breakthroughs in AI make it more accessible — and much more interesting — to users. While tech giants like Microsoft and Google have already introduced their versions of conversational AI tools built using large language models (LLMs), other lesser-known companies have thrown themselves in the mix, setting the stage for an AI showdown.
Here, we’ve rounded up a list of all the companies and AI chatbots that are looking to challenge ChatGPT — or build on top of its success.
Let’s start with Microsoft. The company made its chatbot debut with its launch of the “new” Bing, which promises to upend the way we search for things online. It also built AI-powered tools into the Edge browser.
Microsoft — a big investor in OpenAI — leveraged the technology behind ChatGPT to build an AI tool it says is “even more powerful.” So far, the results have oscillated between impressive and truly off the rails.
The company made the “new” Bing available for beta testers, who have been able to ask questions like “Can you suggest places to visit in Paris?” or “What’s the best apple pie recipe?” and then receive annotated responses describing various tourist destinations or outline the ingredients and steps that go along with a recipe.
But Microsoft may have made Bing a bit too flexible. Users quickly found exploits with the system, including a now-disabled prompt that triggers the Bing bot to divulge its internal nickname, Sydney, and some of the parameters its developers set for its behavior, such as “Sydney’s responses should avoid being vague, controversial, or off-topic.”
Other users toying with the system have found pleasure in pushing the bot’s buttons, triggering wacky — and sometimes unhinged — responses. Microsoft introduced a five-answer limit and a 50-question cap to help curb some of Bing’s more outlandish replies, but the company later loosened some of these restrictions after receiving complaints from users.
As for Edge, Microsoft plans on adding AI enhancements that let you summarize the webpage or document you’re reading online, as well as generate text for social media posts, emails, and more.
Google couldn’t let Microsoft get away with launching an AI chatbot that has the potential to challenge the company’s core business: search. That’s why it rushed to announce its own AI chatbot, Bard, though we still don’t know much about its capabilities.
According to Google CEO Sundar Pichai, the company is using its in-house large language model, LaMDA, to power the conversational AI service, which “draws on information from the web to provide fresh, high-quality responses.” Google says you’ll be able to use the chatbot for a range of tasks, like planning a baby shower, comparing two Oscar-nominated movies, and getting recipe ideas based on the ingredients you have in your fridge.
The company’s announcement was considerably more haphazard than Microsoft’s, so much so that Googlers reportedly criticized the company for it in internal messages. Bard made a factual error in the very first demo Google posted to Twitter, and a presenter showing off the chatbot during a search event in Paris forgot the phone they were supposed to use during the presentation. Bard is currently only available to a limited test group, with wider availability arriving in the “coming weeks.”
Meta — the company that owns Facebook, Instagram, and WhatsApp — also has its sights set on AI. It developed Galactica, a language model designed to provide assistance to scientists and researchers with summaries of academic articles, solutions to math problems, the ability to annotate molecules, and more.
While Meta says it trained the bot on “over 48 million papers, textbooks, reference material, compounds, proteins and other sources of scientific knowledge,” the bot produced disappointing results when the company made it available in a public beta last November. The scientific community fiercely criticized the tool, with one scientist calling it “dangerous” due to its incorrect or biased responses. Meta took the chatbot offline after just a few days.
Galactica isn’t Meta’s first stab at developing an AI model. It also created BlenderBot 3, which is supposed to act like a digital assistant of sorts. Meta made the bot available to the public last August, and it isn’t particularly impressive. When testing the chatbot, Vox’s Kelsey Piper said that its answers “were really poor” and called GPT-3 — the framework that ChatGPT’s built upon — “wildly better” than BlenderBot. BlenderBot 3 is still available online, despite it bad-mouthing Meta CEO Mark Zuckerberg and saying all kinds of offensive things.
There’s more to come from Meta in the AI space just yet. CEO Mark Zuckerberg announced that the company established a dedicated AI team that will eventually create “AI personas” designed to help people, as well as text- and image-based AI tools for WhatsApp, Instagram, and Messenger.
Anthropic, an AI research company founded by former OpenAI employees in 2021, is working on a Chat-GPT competitor of its own called Claude, which has yet to get a full public release. Google invested $300 million into Anthropic in late 2022.
The company developed the chatbot using a methodology it calls Constitutional AI. There’s a whole research paper about the framework here, but, in short, it involves Anthropic training the language model with a set of around 10 “natural language instructions or principles” that it uses to revise its responses automatically. The goal of the system, according to Anthropic, is to “train better and more harmless AI assistants” without incorporating human feedback.
Scale, an AI data platform, was given access to Claude and outlined some of the differences between Anthropic’s bot and Chat-GPT. It found that the service could serve as a “serious” competitor to the OpenAI-made system and that the bot was “more inclined to refuse inappropriate requests.” It does come with some drawbacks, however, as Claude still appears to be prone to making factual errors and mathematical mistakes. For now, the general public can’t access Claude, and it’s only available to companies as an early-access product.
