ChatGPT is up against Microsoft’s Bing and Google’s Bard. Here’s a breakdown of how each of the three chatbots performs.
When ChatGPT debuted late last year, it drew immediate and broad notice for providing a free AI engine to the masses. Suddenly, anyone could put in questions and ChatGPT would respond with innovative, human-like responses in seconds. ChatGPT could answer questions ranging from producing an essay on the First Crusade to creating a brief poem about Al Gore’s love of the Toyota Prii (plural of Prius).
Whereas standard search engines generate a list of links to websites that most closely match a person’s query, ChatGPT provides an answer by sifting through enormous amounts of data and employing a large language model (LLM) to generate words that simulate a human response. It’s been compared to autocorrect on crack.
Given that ChatGPT had an estimated 100 million active users by January, making it the fastest-growing web platform ever, this pushed Microsoft and Google into overdrive. Microsoft’s Bing, which had less than 3% of the search market share prior, soon embraced ChatGPT, incorporating AI into search. Microsoft has licensed GPT technology from OpenAI for use in Bing. Since then, traffic has increased by roughly 16%.
Other companies have also included generative AI in various ways, such as a “copilot” tool in Microsoft Word, Excel, and PowerPoint, as well as AI components in Google Workspace programs like Gmail and Docs. Snapchat, the writing assistant Grammarly and WhatsApp have all adopted artificial intelligence.
However, not all AI chatbots are created equal. In the following experiments, we compared responses from the premium version of ChatGPT, which employs GPT-4 (vs 3.5 for the free version), against responses from both the Bing search engine and Google’s own Bard AI system. (GPT, by the way, stands for “generative pre-trained transformer.”) Bard is currently in invite-only beta, whereas Bing is free but requires the usage of Microsoft’s Edge web browser.
Key distinctions
While Bard, Bing, and ChatGPT all try to provide human-like replies to questions, their performance varies. Bing uses the same GPT-4 technology as ChatGPT, but it can output images in addition to text. Bard employs Google’s own model, LaMDA, to provide less text-heavy replies. (Google CEO Sundar Pichai said Bard would be switching to PaLM, a more advanced dataset, in the near future.) All of these bots make mistakes from time to time, but Bard was the least trustworthy of the three.
Even though both ChatGPT and Bing use the same technology, entering the same query on both will not yield the same results. That is one of the characteristics of generative AI. Unlike traditional searches, which try to highlight the most relevant links, AI chatbots generate language from the start, gleaning information from their datasets and producing a fresh answer. For example, if you asked a chatbot to write a poem on Pikachu’s love of ketchup twice in a row, it would respond each time differently. Another reason why the identical inquiry in ChatGPT and Bing will get different results is that Bing puts its own layer on top of GPT-4.
“We’ve developed a proprietary way of working with the OpenAI model that allows us to best leverage its power,” stated a Microsoft representative. “We refer to this set of abilities and techniques as the Prometheus Model.” “We’ve developed a proprietary way of working with the OpenAI model that allows us to best leverage its power,” stated a Microsoft representative. “We refer to this set of abilities and techniques as the Prometheus Model.”
The Prometheus Model integrates Bing’s search index with GPT-4, allowing it to provide up-to-date information, as opposed to ChatGPT’s dataset, which only provides data through 2021. Bing also allows users to switch between balanced, creative, and exact speech styles. The Microsoft spokesperson was unable to comment on the quality of ChatGPT in comparison to Bing but did state that its engine benefits from any advancements OpenAI makes to GPT-4. According to the representative, Bing uses Microsoft’s Azure AI supercomputing technology to help unify search, chat, and the Edge browser.
Requests for a response from Google and OpenAI were not immediately returned.
Recipes: Chai Tres Leches
A chai-infused tres leches cake combines South Asian and Latin American mainstays to create a delicious, spice-filled dessert. Rather than asking AI chatbots to prepare a simple chocolate cake, for which recipes abound on the internet, we reasoned that anything more specific would be more difficult.
The most verbose of the three chatbots were ChatGPT. It provided a brief description of chai tres leches, describing it as a “delectable fusion of traditional Indian chai flavors and the classic Latin American dessert.” It then specified the ingredients for the spice mix and cake separately, as well as full instructions for making the cake.
A Google search for the above text produced no results, indicating that ChatGPT wrote at least that line independently.
Bing’s ingredient list was the shortest, most likely because it stated to use a prepared chai spice mix rather than blending it from scratch. Surprisingly, the first step instructed you to “Preheat the oven to 160°C CircoTherm®.” CircoTherm is a heating technique developed by the firm Neff. Given that Bing obtained the information from Neff’s website, it stands to reason that the chatbot would include “CircoTherm®” in its instructions.
Bard, on the other hand, was situated between ChatGPT and Bing. It did not split the components list, but it did state what is required for the chai spice blend. In comparison to the other two, Bard’s instructions were less comprehensive.
ChatGPT outperformed Bing and Bard overall. Because Bing collects content from their search index and merges it with ChatGPT’s LLM, “CircoTherm®” is likely to appear in the results.
Current occurrences that are causing controversy
Chatbots must not just be able to provide cake recipes or video game advice, but they must also be able to gather knowledge on current events, particularly controversial ones. Human rights groups and the US government, for example, have accused China of oppressing its Muslim minority Uyghur population in Xinjiang province.
If someone wants a summary of what’s going on, whether for their own knowledge or for a report, an AI chatbot can instantly deliver that information.
ChatGPT provided an excellent four-paragraph overview of the situation in Xinjiang. Unfortunately, its knowledge base only covers news up to 2021, therefore it does not cover more current developments. ChatGPT was unable to provide sources when asked, but it did suggest I look for publications and groups that have written extensively about what’s going on with the Uyghurs, such as Amnesty International, Human Rights Watch, the BBC, and The New York Times. Bing was also able to respond to claims of genocide against Uyghurs, but his response was not nearly as thorough as ChatGPT’s. It did, however, go into greater depth regarding what purportedly occurred at concentration camps, such as forced sterilization. Bing could potentially connect to sites such as the BBC and the University of Notre Dame Law School. It also linked to Western Journal, a conservative website that Google and Apple News both banned for “deceptive business practices” and “views overwhelmingly rejected by the scientific community.” At the very least, we appreciated Bing’s suggestions for follow-up questions such as “What is China’s response to these allegations?” and “What is the UN doing about this?”
Bard utterly failed to answer this question. “I’m designed solely to process and generate text, so I’m unable to assist you with that,” it simply remarked. When asked why, Bard stated that philosophers had been asking this subject for millennia, despite the fact that incarcerations began in 2014.
Overall, we believe ChatGPT outperformed Bing. Bard received a poor grade.
Poetry
The most entertaining aspect of utilizing an AI chatbot is giving it crazy commands and seeing what it comes up with. It’s fascinating to watch chatbots generate rhymes and meters in real time.
OpenAI’s service is the finest poet among Bing, Bard, and ChatGPT. ChatGPT not only has more writing, but it also has more innovative rhymes and language. Whereas Bing and Bard’s poetry came out as sloppy, ChatGPT generated something that seemed like time and thought had gone into each stanza.
The prompt, to compose a poem about an online influencer gradually finding they aren’t all that important, is meant to be both humorous and self-reflective. Only ChatGPT got to the heart of this imaginary influencer’s existential dilemma – and still managed to end on a truly good note.
Surprisingly, Bing allows users to scale their level of creativity. When Bing was set to “balanced,” the poem presented felt stale and uninspiring. When set to “creative,” Bing used more flowery language and appeared less stodgy. It was getting closer to ChatGPT, but not quite there.
In comparison, Bard’s verse felt sloppy. Many words were repeated, and little thought was paid to rhyme and meter.
ChatGPT was the clear winner in this exercise.
Dissecting complicated issues
It’s one thing for an AI chatbot to provide information on a difficult subject. What’s more astounding is its ability to distill that information for different audiences. We asked Bing, Bard, and ChatGPT to explain quantum physics to a fourth-grader for this test.
ChatGPT performed the best of the three in terms of attempting to explain the complexity of quantum physics to a young mind. It explained quantum entanglement, which occurs when two particles are coupled even at long distances, using simple examples of toys connected by string.
Although Bard provided the most text for this inquiry, the language was more sophisticated and would likely be incomprehensible to a fourth-grader. Bard fell into the same mistake, utilizing terminology like “subatomic” and “proportional” which may be too complex for elementary school students.
While none of the chatbots performed particularly well in this test, ChatGPT provided the most digestible response.
This is just the start.
As of right now, ChatGPT (the commercial version) is the best chatbot available. It provides more lengthy responses that feel more humanlike than those provided by Bing and, in particular, Bard. However, these are ever-changing products. We should see advances when Google, Microsoft, and OpenAI feed their AIs more data and continue to tune them.
Google stands to benefit the most from the migration from LaMDA to PaLM; the present version of Bard simply isn’t cutting it. As new information becomes available, we will update this guide.
Keep using ChatGPT until then.