Given all the buzz around ChatGPT, it makes sense that other businesses would want to get in on the AI-powered chatbot action. Businesses believe that the artificial intelligence market is at a turning point, where goods that adopt and build upon the emerging technology may have the capacity to change technology as we know it and upend the Big Tech hierarchy.
The stakes are enormous, and the greatest names in technology don’t want to fall behind as advances in AI make it more approachable and engaging for users. While well-known software companies like Microsoft and Google have already debuted conversational AI capabilities made with LLMs, other less well-known businesses have entered the fray, setting the stage for an AI competition.
Microsoft
Start with Microsoft first. With the introduction of the “new” Bing, which is expected to revolutionize the way we conduct internet searches, the firm made its chatbot premiere. Moreover, it included Edge’s browser with AI-powered capabilities.
Microsoft, a significant OpenAI investor, used ChatGPT’s technology to create a “far more powerful” AI tool. The outcomes thus far have fluctuated between being totally off the rails and being outstanding.
The business let beta testers to access the “new” Bing and ask inquiries like, “Can you recommend places to visit in Paris?” or “What’s the greatest apple pie recipe?” and get annotated answers that provide the ingredients and instructions for a recipe or describe various tourist locations.
But Bing may have been overly adaptable thanks to Microsoft. The Bing bot’s internal nickname, Sydney, and some of the guidelines its creators set for its behavior, such as “Sydney’s responses should avoid being unclear, contentious, or off-topic,” were rapidly discovered to be flaws by users. This prompt is now blocked.
Other users who were playing around with the system enjoyed pressing the bot’s buttons, which produced bizarre — and occasionally insane — reactions. To help rein in some of Bing’s most bizarre responses, Microsoft imposed a five-answer limit and a 50-question maximum. Nevertheless, the company later relaxed some of these restrictions in response to customer complaints.
Microsoft intends to improve Edge with AI features that will enable you to sum up the webpage or document you’re reading online and create text for emails, social media posts, and other purposes.
Microsoft couldn’t get away with releasing an AI chatbot that could threaten Google‘s primary revenue stream, search. In spite of the fact that we still don’t fully understand Bard’s capabilities, it hurried to proclaim it as its own AI chatbot.
The conversational AI service, which “draws on information from the web to give fresh, high-quality responses,” is powered by LaMDA, the company’s internal big language model, according to Google CEO Sundar Pichai. The chatbot will reportedly be useful for a variety of activities, including organizing a baby shower, contrasting two Oscar-nominated films, and finding recipes based on the goods you already have in your refrigerator, according to Google.
The company’s announcement was far more chaotic than Microsoft’s, to the point where Googlers apparently expressed their displeasure with it in internal messages. In the very first demo Google shared to Twitter, Bard made a factual error, and a presenter showcasing the chatbot at a search event in Paris forgot the phone they were supposed to use during the presentation. With wider distribution coming in the “coming weeks,” Bard is presently only accessible to a small test group.
Meta
AI is a focus for Meta, the business that owns Facebook, Instagram, and WhatsApp. It created the language model known as Galactica, which enables scientists and researchers to annotate molecules, summarize academic articles, solve math problems, and more.
The bot was trained on “over 48 million publications, textbooks, reference material, chemicals, proteins and other sources of scientific information,” according to Meta, but when the business made it available in a public beta last November, it performed poorly. The instrument received harsh criticism from the scientific community; one expert even dubbed it “hazardous” because of its inaccurate or biased conclusions. In a short period of time, Meta removed the chatbot.
Galactica is not Meta’s first attempt at creating an AI simulation. It also produced BlenderBot 3, which is designed to function somewhat like a personal assistant. Meta made the bot available to the public in August, and it isn’t particularly impressive. Vox’s Kelsey Piper tested the chatbot and found that its responses “were extremely terrible” but that GPT-3, the framework on which ChatGPT is based, is “wildly better” than BlenderBot. Even if BlenderBot 3 insults Meta CEO Mark Zuckerberg and says a variety of terrible things, it is still accessible online.
But, Meta has more to offer in the field of AI. A dedicated AI team has been established by the firm, according to CEO Mark Zuckerberg. This team will eventually produce “AI avatars” intended to assist users as well as text- and image-based AI capabilities for WhatsApp, Instagram, and Messenger.
Anthropic
Anthropic, an AI research firm established in 2021 by former OpenAI personnel, is developing Claude, a Chat-GPT rival that hasn’t been fully released to the public yet. In the latter half of 2022, Google spent $300 million in Anthropic.
The chatbot was created by the corporation using a process it calls Constitutional AI. The framework is the subject of an entire research paper, but, in brief, it entails Anthropic training the language model with a collection of roughly 10 “natural language commands or principles” that it employs to autonomously revise its responses. Anthropic claims that the system’s objective is to “train better and more harmless AI assistants” without using feedback from people.
When provided access to Claude, Scale, an AI data platform, explained some of the variations between Anthropic’s bot and Chat-GPT. The system created by OpenAI could face “severe” competition from the service, and it was discovered that the bot was “more disposed to decline incorrect requests.” Yet, there are some limitations because Claude still seems to be prone to factual and mathematical errors. Claude is currently only accessible to businesses and is not yet available to the general public.
Alibaba
The large Chinese e-commerce company Alibaba has also embraced the AI chatbot craze. A corporate representative told CNBC in early February that the company is internally testing a Chat-GPT alternative. Since 2017, Alibaba is said to have been experimenting with generative AI, but the business hasn’t indicated when it would reveal the tool it’s developing or what it could be able to do.
Alibaba might have to overcome certain challenges before starting its own ChatGPT, though. According to a Nikkei Asia story, Chinese regulators have already advised Tencent and Ant Group, which are owned by Alibaba, to censor access to ChatGPT due to worries that the bot may promote restricted content. Before releasing their own bots to the public, the corporations must also consult with the government.
All other Chinese businesses creating AI chatbots will likely be subject to similar regulations, raising the possibility that their products won’t ever be able to be released or that China’s stringent censorship regulations may limit their usefulness.
Baidu
Baidu, a different Chinese corporation, is getting ready to introduce “Ernie Bot,” an AI tool, as soon as March. Together with a number of other web-related services like the mapping platform Baidu Maps, the online encyclopedia Baidu Baike, the cloud storage service Baidu Wangpan, and others, Baidu is best known for its search engine of the same name. Also, it is using AI technology to create a self-driving vehicle.
Since its initial release in 2019, Ernie—which stands for Enhanced Representation through kNowledge IntEgration—has developed into a program that resembles ChatGPT and can provide conversational answers. Late in 2021, Baidu claimed that the model “excels in both natural language understanding (NLU) and generation (NLG)” and that it had been trained on “huge unstructured data and a vast knowledge graph.”
Baidu plans to incorporate the chatbot into its search engine, just like Microsoft and Google have done, and will even incorporate the technology into the next electric vehicle produced by Chinese startup Jidu. Baidu is working on a text-to-image model called Ernie ViLG to produce images based on Chinese language in addition to its Chat-GPT-style tool. This model is comparable to OpenAI’s DALL-E 2 system and Stability AI’s Stable Diffusion’s AI image generator.
