Have you ever found yourself sifting through several pages of search engine results to answer a difficult question? Say you want to find out if a vegetarian diet is suitable for your dog. Your research adventure could begin by putting “is a vegetarian diet good for dogs” into Google’s search box and then deciphering the plethora of resulting links. By the time you uncover an answer, you’ve buried considerably more time than you’d budgeted into reading through publications, reports and their sources.
In the not-too-distant future, finding the answer to a difficult issue may no longer be a lengthy, mind-numbing procedure. Microsoft is apparently incorporating the artificial intelligence technology that underpins ChatGPT into its Bing search engine, a move that could fundamentally alter the search industry. According to AI researchers, Bing may have the capacity to provide a search experience that is superior to Google, so challenging the search giant’s decades-long supremacy.
“ChatGPT is the first new technology in more than a decade that has the potential to truly alter search and that could, at least in theory, challenge Google’s market supremacy,” said University of Virginia AI researcher and economics professor Anton Korinek. The technology enables users to engage with their computer in a manner that is far more natural and conversational than standard search.
The actual appearance of Bing’s AI-driven search results is yet unknown. Microsoft refused to comment on this article. In terms of how a search engine delivers an answer and how users interact with it, AI experts anticipate a significant change from the current quo. After all, ChatGPT is not designed to scan the internet for information (like a search engine) (like a search engine). Instead, the chatbot generates responses using information gleaned from large amounts of training data.
“ChatGPT can provide its users with a single unambiguous response compared to the myriads of links of standard search engines. It also has features that are much beyond standard search engines, such [the capacity] to generate new text, explain concepts, have a back-and-forth discussion between the user and the system, and so on,” added Korinek. “People still find emergent capabilities that even the creators of ChatGPT were not aware that the system had.”
Microsoft stated on January 23 that it will invest $10 billion in OpenAI, the company that created ChatGPT. The merger would help maintain both companies at the cutting edge of what’s known as generative AI, a tech utilised in ChatGPT that can learn from huge amounts of data to create virtually any content type (writing, photos, audio and so on) just from a text input.
Search is only one of a range of consumer-facing Microsoft products that could undergo significant transformations in the next years. According to a report by The Information, the Seattle-based tech giant also has plans to integrate ChatGPT’s artificial intelligence (AI) technology into well-established products such as Word, PowerPoint, and Outlook, a move that could alter the way over a billion people work and complete daily tasks. For example, integrating it with Outlook could involve instructing the email application to compose a message about a particular topic.
“Microsoft will deploy OpenAI’s models across our consumer and enterprise products and introduce new categories of digital experiences built on OpenAI’s technology,” the company said in a news release announcing the cooperation expansion.
Google and its innovative subsidiary DeepMind have been developing such systems for years. However, the search engine giant elected not to reveal them due in part to worries about unethical activity and how chat platforms can often violate societal standards. Microsoft, for instance, developed a chatbot named Tay in 2016 that was taken down after spewing hate remarks. Even ChatGPT, which includes guidelines to promote good and friendly content, may be fooled into producing distressing responses utilising the correct cues.
In contrast, DeepMind CEO Demis Hassabis stated in an interview with Time this month that the company is exploring a “private beta” release of its own chatbot, Sparrow, in 2023. Google aims to demonstrate a version of its search engine with chatbot capabilities as early as this year, according to a report from The New York Times citing persons with knowledge of the matter.
It is no secret that Google search has gotten increasingly conversational over time. The business has made inroads in this area with the Google assistant and knowledge panels in search, and has touted conversation as the future of search for years, demonstrating LaMDA and MUM at its 2021 I/O developer conference.
Utilizing OpenAI’s artificial intelligence appears to be Microsoft’s strategy for competing with Google at its own game. According to The New York Times, Google management issued a “code red” in response to the release of ChatGPT. The article said internal teams have been reassigned to commence work on AI between now and a planned business meeting in May.
According to Statista, Google’s search engine will hold 84% of the global search market share in 2022, compared to Bing’s 9% share (although Bing’s market share has increased in recent years).
Google did not reply promptly to a request for comment on this article.
How smart is ChatGPT?
As you’ve surely heard by now, ChatGPT is a smart chatbot that went popular worldwide upon its introduction to consumers in late November as a free web application accessible to anybody with an internet connection. The chatbot driven by artificial intelligence garnered headlines in part because to its capacity to produce charming poetry, prepare meal plans, and deliver authoritative answers to hard inquiries within seconds of being requested. Its underlying technology is not exactly cutting-edge, but before to ChatGPT, no chatbot had succeeded to grab the interest of the general public. This is largely due to the fact that OpenAI designed a slick user interface for the GPT-3.5 language model, a phenomena known as ChatGPT.
GPT-3.5 is an upgraded version of GPT-3, which launched in 2020 and utilised enormous amounts of data and code to acquire its capabilities. According to Stanford University researchers, GPT-3 was trained using 570 terabytes of text and includes 175 billion parameters. According to Google’s Dale Markowitz, there are 45 terabytes of text data, “containing nearly the whole public web.” Comparatively, its predecessor, GPT-2, has 1.5 billion parameters, over 100 times fewer.
“This increase in scale substantially alters the model’s behaviour, allowing GPT-3 to accomplish tasks on which it was not expressly taught, such as translating words from English to French, with very few or no training samples. This tendency was mostly missing in GPT-2,” researchers from Stanford’s Institute for Human-Centered Artificial Intelligence stated in a 2021 post.
“The current version of ChatGPT likely knows more about the world than any single human, and it can deliver this information in an easily digestible format,” added Korinek.
For all the promise ChatGPT holds, there are nearly as many constraints. Critics of ChatGPT assert that it is not always evident where the chatbot obtains its information, making it difficult for individuals to trust the results. Skeptics further assert that ChatGPT will always be hampered by the defective nature of the data on which it was trained, such as biassed or false information.
OpenAI has acknowledged the chatbot’s current shortcomings. In a December tweet, CEO Sam Altman stated that the product lacks “robustness and veracity” and that it would be unwise to rely on it for anything significant at this time.
“In 2023, a number of new systems similar to ChatGPT will hit the market, and the main meaning of the subsequent rivalry will be that consumers will have more options and, ideally, better products,” explained Korinek.
ChatGPT is incredibly limited, but good enough at some things to create a misleading impression of greatness.
it's a mistake to be relying on it for anything important right now. it’s a preview of progress; we have lots of work to do on robustness and truthfulness.
— Sam Altman (@sama) December 11, 2022
According to reports, the under-development GPT-4 has 100 trillion parameters. However, a release is not anticipated until OpenAI is “sure we can distribute it securely and responsibly,” as Altman told StrictlyVC in early January.
Altman has attempted to temper expectations for the fourth version of GPT, the complex language model underlying ChatGPT, by stating, “we do not have AGI.” AGI refers to artificial general intelligence, or a technology having its own emergent intelligence as opposed to OpenAI’s current reliance on deep learning models. It’s the kind of intelligence that has been dramatised in science fiction stories for more than a century and was popularised in recent years by the award-winning dystopian show Westworld.
“I think [AGI] is sort of what is anticipated of us,” Altman said in the same interview, adding that GPT-4 is “going to disappoint” anyone who hold out that hope.