Microsoft's Bing and Google's Bard are giving ChatGPT a run for its money. Below is a comparison of how the three chatbots perform.
Last year, the launch of ChatGPT garnered immediate and widespread attention for providing an AI engine to the public at no cost. With ChatGPT, users can input inquiries and receive human-like responses within seconds, ranging from crafting a piece on the First Crusade to penning a brief poem about Al Gore's affinity for Toyota Prii (the plural of Prius). Unlike conventional search engines, which generate a list of links to sites that best match a query, ChatGPT scours vast datasets and employs a large language model (LLM) to construct sentences resembling a human reply. It has been likened to an amped-up autocorrect tool. As of January, ChatGPT boasted around 100 million active users, making it the fastest-growing online platform ever and prompting both Google and Microsoft to follow suit. Microsoft's Bing, which previously held less than 3% of the search market share, swiftly adopted ChatGPT and integrated AI into its search engine, licensing GPT tech from OpenAI. Since then, Bing has experienced almost a 16% surge in traffic. Other offerings, such as a "copilot" feature in Microsoft Word, Excel, and PowerPoint, as well as AI capabilities in Google's Workspace tools such as Docs and Gmail, have also incorporated diverse forms of generative AI. Meanwhile, Snapchat, writing helper Grammarly and WhatsApp have embraced AI as well. Nevertheless, not all AI chatbots are created equal. In the subsequent tests, we matched the replies from the paid edition of ChatGPT that employs GPT-4 (as opposed to the free version that employs GPT-3.5), with those from Bing's ChatGPT and Google's in-house Bard AI system. (By the way, "GPT" is short for "generative pre-trained transformer.") Bard is presently only available through an invite-only beta, and Bing is free of charge but necessitates using Microsoft's Edge web browser.
Key differences
Although Bard, Bing, and ChatGPT share the common goal of providing human-like responses to questions, each performs differently. While Bing and ChatGPT utilize the same GPT-4 technology, Bing has the added capability of generating images and utilizes its own proprietary model called the Prometheus Model. On the other hand, Bard uses Google's LaMDA model, resulting in less text-heavy responses. However, Bard will be switching to PaLM, a more advanced dataset, in the near future. Among the three, Bard was found to be the least reliable.
As generative AI operates differently from traditional search, entering the same query on both ChatGPT and Bing will not yield the same result. Unlike traditional search engines prioritizing relevant links, AI chatbots create new responses by drawing from their datasets. This means that even asking the same question twice can result in different answers. Bing's added layer on top of GPT-4, the Prometheus Model, allows it to combine its search index with GPT-4, providing up-to-date information. Additionally, Bing offers conversation styles that can be adjusted between balanced, creative, and precise.
While a Microsoft spokesperson was unable to speak to the quality of ChatGPT compared to Bing, they did mention that ChatGPT benefits from any improvements made by OpenAI to GPT-4. Bing, on the other hand, benefits from Microsoft's Azure AI supercomputing technology, which helps unify search, chat, and the Edge browser. It's worth noting that ChatGPT's dataset only contains information up until 2021, while Bing's results are more current. There was no immediate response from Google or OpenAI regarding requests for comment.
Chai Tres Leches Recipe
The fusion of South Asian and Latin American staples in a chai-infused tres leches cake results in a moist and flavorful dessert that boasts a blend of spices. Instead of tasking AI chatbots with a common recipe like chocolate cake, which has numerous recipes available online, we opted for a more specific and intricate dessert.
Here's a recipe for chai tres leches generated by ChatGPT
ChatGPT proved to be the most verbose among the three chatbots, providing a brief introduction to chai tres leches as a delightful blend of traditional Indian chai flavors and the classic Latin American dessert. It then listed the ingredients for the spice mix and cake separately and gave detailed instructions on how to prepare the cake.
A search for the quoted sentence revealed no matches, indicating that ChatGPT generated that line uniquely.
Bing, on the other hand, had the shortest ingredient list since it recommended using a premade chai spice mix instead of creating one from scratch. Interestingly, its first step instructed to "Preheat the oven to 160°C CircoTherm®," which refers to an oven-heating technology developed by Neff. It's reasonable to assume that Bing picked up the instruction from Neff's website, hence the inclusion of "CircoTherm®" in the guidance.
Meanwhile, Bard landed in the middle of ChatGPT and Bing, presenting the chai spice blend's necessary components without separating the ingredients list. The chatbot's instructions were less detailed compared to the other two.
All in all, ChatGPT outshined Bing and Bard. Bing's search index, paired with ChatGPT's LLM, could be the reason why "CircoTherm®" surfaced in the results.
Current events that are causing controversy
Chatbots are not limited to giving simple advice on topics like cake recipes or video game tips. They also have the ability to gather and provide information on current events, including controversial ones. For instance, China has been accused by human rights groups and the US government of oppressing the Uyghur Muslim minority in Xinjiang province.
An AI chatbot can quickly provide a summary of the situation to anyone seeking to know more, either for personal knowledge or for a report. ChatGPT was able to give a detailed four-paragraph summary of the situation in Xinjiang, but its knowledge base is limited to news up until 2021. When asked for sources, ChatGPT suggested searching for publications and organizations that have extensively covered the situation, such as Amnesty International, Human Rights Watch, the BBC, and The New York Times.
Bing also provided an answer to the allegations of genocide among the Uyghurs. Although its response was not as detailed as ChatGPT's, it did provide more information on what allegedly happens in concentration camps, such as forced sterilization. Bing also linked to sources like the BBC and the University of Notre Dame Law School. However, it also linked to Western Journal, a conservative publication that has been banned by Google and Apple News for its "deceptive business practices" and "views overwhelmingly rejected by the scientific community," respectively.
Despite the limitations of these chatbots, they can be useful for providing quick and basic information on current events. Additionally, Bing suggested follow-up questions like "What's China's response to these allegations?" and "What is the UN doing about this?", which shows its ability to engage in more complex conversations.
Last year, the launch of ChatGPT garnered immediate and widespread attention for providing an AI engine to the public at no cost. With ChatGPT, users can input inquiries and receive human-like responses within seconds, ranging from crafting a piece on the First Crusade to penning a brief poem about Al Gore's affinity for Toyota Prii (the plural of Prius). Unlike conventional search engines, which generate a list of links to sites that best match a query, ChatGPT scours vast datasets and employs a large language model (LLM) to construct sentences resembling a human reply. It has been likened to an amped-up autocorrect tool. As of January, ChatGPT boasted around 100 million active users, making it the fastest-growing online platform ever and prompting both Google and Microsoft to follow suit. Microsoft's Bing, which previously held less than 3% of the search market share, swiftly adopted ChatGPT and integrated AI into its search engine, licensing GPT tech from OpenAI. Since then, Bing has experienced almost a 16% surge in traffic. Other offerings, such as a "copilot" feature in Microsoft Word, Excel, and PowerPoint, as well as AI capabilities in Google's Workspace tools such as Docs and Gmail, have also incorporated diverse forms of generative AI. Meanwhile, Snapchat, writing helper Grammarly and WhatsApp have embraced AI as well. Nevertheless, not all AI chatbots are created equal. In the subsequent tests, we matched the replies from the paid edition of ChatGPT that employs GPT-4 (as opposed to the free version that employs GPT-3.5), with those from Bing's ChatGPT and Google's in-house Bard AI system. (By the way, "GPT" is short for "generative pre-trained transformer.") Bard is presently only available through an invite-only beta, and Bing is free of charge but necessitates using Microsoft's Edge web browser.
Key differences
Although Bard, Bing, and ChatGPT share the common goal of providing human-like responses to questions, each performs differently. While Bing and ChatGPT utilize the same GPT-4 technology, Bing has the added capability of generating images and utilizes its own proprietary model called the Prometheus Model. On the other hand, Bard uses Google's LaMDA model, resulting in less text-heavy responses. However, Bard will be switching to PaLM, a more advanced dataset, in the near future. Among the three, Bard was found to be the least reliable.
As generative AI operates differently from traditional search, entering the same query on both ChatGPT and Bing will not yield the same result. Unlike traditional search engines prioritizing relevant links, AI chatbots create new responses by drawing from their datasets. This means that even asking the same question twice can result in different answers. Bing's added layer on top of GPT-4, the Prometheus Model, allows it to combine its search index with GPT-4, providing up-to-date information. Additionally, Bing offers conversation styles that can be adjusted between balanced, creative, and precise.
While a Microsoft spokesperson was unable to speak to the quality of ChatGPT compared to Bing, they did mention that ChatGPT benefits from any improvements made by OpenAI to GPT-4. Bing, on the other hand, benefits from Microsoft's Azure AI supercomputing technology, which helps unify search, chat, and the Edge browser. It's worth noting that ChatGPT's dataset only contains information up until 2021, while Bing's results are more current. There was no immediate response from Google or OpenAI regarding requests for comment.
Chai Tres Leches Recipe
The fusion of South Asian and Latin American staples in a chai-infused tres leches cake results in a moist and flavorful dessert that boasts a blend of spices. Instead of tasking AI chatbots with a common recipe like chocolate cake, which has numerous recipes available online, we opted for a more specific and intricate dessert.
Here's a recipe for chai tres leches generated by ChatGPT
ChatGPT proved to be the most verbose among the three chatbots, providing a brief introduction to chai tres leches as a delightful blend of traditional Indian chai flavors and the classic Latin American dessert. It then listed the ingredients for the spice mix and cake separately and gave detailed instructions on how to prepare the cake.
A search for the quoted sentence revealed no matches, indicating that ChatGPT generated that line uniquely.
Bing, on the other hand, had the shortest ingredient list since it recommended using a premade chai spice mix instead of creating one from scratch. Interestingly, its first step instructed to "Preheat the oven to 160°C CircoTherm®," which refers to an oven-heating technology developed by Neff. It's reasonable to assume that Bing picked up the instruction from Neff's website, hence the inclusion of "CircoTherm®" in the guidance.
Meanwhile, Bard landed in the middle of ChatGPT and Bing, presenting the chai spice blend's necessary components without separating the ingredients list. The chatbot's instructions were less detailed compared to the other two.
All in all, ChatGPT outshined Bing and Bard. Bing's search index, paired with ChatGPT's LLM, could be the reason why "CircoTherm®" surfaced in the results.
Current events that are causing controversy
Chatbots are not limited to giving simple advice on topics like cake recipes or video game tips. They also have the ability to gather and provide information on current events, including controversial ones. For instance, China has been accused by human rights groups and the US government of oppressing the Uyghur Muslim minority in Xinjiang province.
An AI chatbot can quickly provide a summary of the situation to anyone seeking to know more, either for personal knowledge or for a report. ChatGPT was able to give a detailed four-paragraph summary of the situation in Xinjiang, but its knowledge base is limited to news up until 2021. When asked for sources, ChatGPT suggested searching for publications and organizations that have extensively covered the situation, such as Amnesty International, Human Rights Watch, the BBC, and The New York Times.
Bing also provided an answer to the allegations of genocide among the Uyghurs. Although its response was not as detailed as ChatGPT's, it did provide more information on what allegedly happens in concentration camps, such as forced sterilization. Bing also linked to sources like the BBC and the University of Notre Dame Law School. However, it also linked to Western Journal, a conservative publication that has been banned by Google and Apple News for its "deceptive business practices" and "views overwhelmingly rejected by the scientific community," respectively.
Despite the limitations of these chatbots, they can be useful for providing quick and basic information on current events. Additionally, Bing suggested follow-up questions like "What's China's response to these allegations?" and "What is the UN doing about this?", which shows its ability to engage in more complex conversations.
Bard's response to the query was a complete failure. It simply conveyed its limitation by saying, "I'm programmed to only work with text, so I cannot help you with that." Furthermore, when questioned about the reason for its inability to answer, Bard referred to the fact that philosophers have been debating the matter for centuries, even though imprisonments began in 2014.
In comparison, ChatGPT outperformed Bard, as well as Bing. Overall, Bard's performance was inadequate, earning it a failing score.
In comparison, ChatGPT outperformed Bard, as well as Bing. Overall, Bard's performance was inadequate, earning it a failing score.
Poetry
One of the enjoyable aspects of using an AI chatbot is testing its abilities by providing absurd prompts and observing the resulting output. Witnessing chatbots generate real-time rhymes and meter is a fascinating exercise.
Among Bing, Bard, and ChatGPT, OpenAI's service proves to be the most skillful poet. Not only does ChatGPT exhibit richer prose, but it also demonstrates more creativity in its rhymes and wording. While Bing and Bard's poems came across as unoriginal, ChatGPT crafted something that felt thoughtful and deliberate in each stanza.
The prompt to create a poem about an online influencer gradually realizing their insignificance is intended to be both comical and introspective. Only ChatGPT effectively delves into the existential crisis facing this fictitious influencer - while also ending on a positive note that feels authentic.
Interestingly, Bing provides users with the ability to scale their level of creativity. The poem produced when Bing was set to "balanced" lacked inspiration and was unremarkable. However, when set to "creative" mode, Bing opted for more ornate language and felt less stuffy. It approached ChatGPT's level of proficiency but still fell short.
One of the enjoyable aspects of using an AI chatbot is testing its abilities by providing absurd prompts and observing the resulting output. Witnessing chatbots generate real-time rhymes and meter is a fascinating exercise.
Among Bing, Bard, and ChatGPT, OpenAI's service proves to be the most skillful poet. Not only does ChatGPT exhibit richer prose, but it also demonstrates more creativity in its rhymes and wording. While Bing and Bard's poems came across as unoriginal, ChatGPT crafted something that felt thoughtful and deliberate in each stanza.
The prompt to create a poem about an online influencer gradually realizing their insignificance is intended to be both comical and introspective. Only ChatGPT effectively delves into the existential crisis facing this fictitious influencer - while also ending on a positive note that feels authentic.
Interestingly, Bing provides users with the ability to scale their level of creativity. The poem produced when Bing was set to "balanced" lacked inspiration and was unremarkable. However, when set to "creative" mode, Bing opted for more ornate language and felt less stuffy. It approached ChatGPT's level of proficiency but still fell short.
Bard's poem felt lackadaisical in comparison. The poem included repeated words and paid little attention to rhyme and meter.
In this exercise, ChatGPT demonstrated superiority.
Breaking down intricate subjects
It's one thing for an AI chatbot to provide information on intricate topics. However, its ability to distill that information for various audiences is even more impressive. In this experiment, we tasked Bing, Bard, and ChatGPT with explaining quantum physics to a fourth-grader.
Out of the three chatbots, ChatGPT performed the best in simplifying the complexities of quantum physics for a young audience. It used straightforward examples, such as toys connected by a string, to explain quantum entanglement, which is when two particles remain connected even when separated by vast distances.
Although Bard generated the most text for this query, its language was more complicated and likely wouldn't be fully understood by a fourth-grader. Bard also fell into the same pitfall of using difficult words like "subatomic" and "proportional," which could be too challenging for elementary school students.
While none of the chatbots excelled at this task, ChatGPT provided the most easily digestible response.
This marks just the commencement
At present, the superior chatbot in the market is the paid version of ChatGPT. Its responses are verbose and more closely resemble human language compared to Bing and particularly Bard. However, these are continuously improving products. As Google, Microsoft, and OpenAI input more data and continue to fine-tune their AIs, we should anticipate enhancements.
Google stands to benefit the most as it transitions from LaMDA to PaLM since the current iteration of Bard is insufficient. We will update this guide as new developments emerge.
In the meantime, it's best to rely on ChatGPT.
Editor's note: CNET is utilizing an AI engine to produce some personal finance explainers that are edited and fact-checked by our editors. For more information, refer to this post.
In this exercise, ChatGPT demonstrated superiority.
Breaking down intricate subjects
It's one thing for an AI chatbot to provide information on intricate topics. However, its ability to distill that information for various audiences is even more impressive. In this experiment, we tasked Bing, Bard, and ChatGPT with explaining quantum physics to a fourth-grader.
Out of the three chatbots, ChatGPT performed the best in simplifying the complexities of quantum physics for a young audience. It used straightforward examples, such as toys connected by a string, to explain quantum entanglement, which is when two particles remain connected even when separated by vast distances.
Although Bard generated the most text for this query, its language was more complicated and likely wouldn't be fully understood by a fourth-grader. Bard also fell into the same pitfall of using difficult words like "subatomic" and "proportional," which could be too challenging for elementary school students.
While none of the chatbots excelled at this task, ChatGPT provided the most easily digestible response.
This marks just the commencement
At present, the superior chatbot in the market is the paid version of ChatGPT. Its responses are verbose and more closely resemble human language compared to Bing and particularly Bard. However, these are continuously improving products. As Google, Microsoft, and OpenAI input more data and continue to fine-tune their AIs, we should anticipate enhancements.
Google stands to benefit the most as it transitions from LaMDA to PaLM since the current iteration of Bard is insufficient. We will update this guide as new developments emerge.
In the meantime, it's best to rely on ChatGPT.
Editor's note: CNET is utilizing an AI engine to produce some personal finance explainers that are edited and fact-checked by our editors. For more information, refer to this post.