Companies Racing to Match ChatGPT's Capabilities

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

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

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

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.

You.com

You.com, a company built by two former Salesforce employees, bills itself as the “search engine you control.” At first glance, it may seem like your typical search engine, but it comes with an AI-powered “chat” tool that works much like the one Microsoft’s piloting on Bing.

You.com first introduced the chatbot, called YouChat, in December 2022 and says it’s built on the company’s C-A-L model, which is “blended with AI-powered conversations, You.com apps, web links and citations.” Just like Microsoft’s AI, YouChat can provide annotated answers to various types of queries, create summaries of articles from the web, generate code, write essays, and more.

In addition to giving users access to an AI-powered chatbot, You.com recently added built-in AI image generator models, including Stable Diffusion 1.5, Stable Diffusion 2.1, and Open Journey, that you can use to generate images based on a written description. The engine also breaks down your search results based on relevant responses on sites like Reddit, TripAdvisor, Wikipedia, and YouTube while also providing standard results from the web.

Alibaba

Alibaba, the China-based e-commerce giant, has caught onto the AI chatbot trend as well. In early February, a company spokesperson told CNBC that the company is testing a Chat-GPT rival internally. Alibaba has reportedly been experimenting with generative AI since 2017, but the company hasn’t provided any sense of when it could announce the tool it’s working on or what it might be capable of.

Alibaba may have to overcome some hurdles before it gets its own version of ChatGPT off the ground, however. A report from Nikkei Asia indicates that Chinese regulators have already told the Alibaba-owned Tencent and Ant Group that they should restrict access to ChatGPT over concerns the bot could espouse uncensored content. The companies will also have to confer with the government before making their own bots available to the public.

Similar rules will likely apply to all of the other Chinese companies developing AI chatbots, calling into question whether they'll even be able to launch their products or if their utility will be held back by China’s strict censorship rules.

Baidu

Another Chinese company, Baidu, is getting ready to launch an AI tool it calls “Ernie Bot” as soon as March. Baidu is best known for its search engine of the same name, along with a cluster of other internet-related services, such as mapping platform Baidu Maps, online encyclopedia Baidu Baike, cloud storage service Baidu Wangpan, and more. It’s also leveraging AI technology to develop a self-driving car.

Ernie, which stands for Enhanced Representation through kNowledge IntEgration, first appeared in 2019 and has since evolved into a ChatGPT-like tool that can generate conversational responses. In late 2021, Baidu said it trained the model on “massive unstructured data and a gigantic knowledge graph” and that it “excels at both natural language understanding (NLU) and generation (NLG).”

Just like Microsoft and Google, Baidu is also planning to integrate the chatbot into its search engine and will even build the tool into the interface of the forthcoming electric vehicle made by Chinese startup Jidu. In addition to this Chat-GPT-style tool, Baidu is also developing a text-to-image model, called Ernie ViLG, to create images based on Chinese text, similar to OpenAI’s DALL-E 2 system and Stability AI’s Stable Diffusion’s AI image generator.

Other possible contenders

Aside from companies making standalone chatbots for search, there are a few other companies using generative AI in slightly different ways.

Snapchat, for example, is working on a “My AI” chatbot that essentially works as an in-app version of ChatGPT, allowing users to ask for recipe suggestions or plan trips. It’s more limiting than ChatGPT, however, as it’s been trained to avoid breaking Snapchat’s trust and safety guidelines. The service is only available as part of Snapchat’s $3.99 per month Plus subscribers for now, but CEO Evan Spiegel plans on eventually bringing it to all users.

Character.AI is another one of these tools and comes from the developers of Google’s LaMDA technology. The site lets you create or browse chatbots modeled after real people or fictional characters, such as Elon Musk, Mark Zuckerberg, or Tony Stark. When “conversing” with these bots, the AI attempts to respond in a manner similar to that person or character’s personality. However, that’s not the only thing these bots are capable of, as some are designed to help generate book recommendations, brainstorm ideas, practice a new language, and more.

Meanwhile, Chinese gaming firm NetEase has announced that its education subsidiary, Youdao, is planning to incorporate AI-powered tools into some of its educational products, according to a report from CNBC. It’s still not clear what exactly this tool will do, but it seems the company’s interested in employing the technology in one of its upcoming games as well.

Daniel Ahmad, the director of research and insights at Niko Partners, reports that NetEase could bring a ChatGPT-style tool to the mobile MMO Justice Online Mobile. As noted by Ahmad, the tool will “allow players to chat with NPCs and have them react in unique ways that impact the game” through text or voice inputs. However, there’s only one demo of the tool so far, so we don’t know how (or if) it will make its way into the final version of the game.

Then, there’s Replika, an AI chatbot that functions as a sort of “companion” that you can talk to via text-based chats and even video calls. The tool combines the company’s own version of the GPT-3 model and scripted dialogue content to build memories and generate responses tailored to your conversation style. But the company that owns the tool recently ruled out erotic roleplay, devastating dedicated users.

We’re still at the beginning of what conversational AI can do, and with major players like Microsoft and Google getting on board, we’re bound to see some progress. It’ll be interesting to see how all of these tools evolve over the coming years, as well as which ones manage to make their way into our daily lives.