Google vs. Microsoft. How Google completely loses to one ChatGPT

Short description

Microsoft is currently leading the race to create the first chatbot that can accurately answer all questions from users. The company invested in OpenAI back in 2019, which has proven to be one of its best investments. However, other tech giants, including Google and Baidu, are working to create their own similar chatbots. Microsoft’s ChatGPT has already attracted over 100 million active users and seen a 100-fold audience growth in just two months. Google recently announced the upcoming release of its chatbot named Bard, although its public presentation failed, causing Alphabet to lose $100 billion in market value.

Google vs. Microsoft. How Google completely loses to one ChatGPT

Microsoft, Google, Baidu and others are working to be the first to create a bot that can answer all your questions. The battle is not for life, but for death. And so far Microsoft is winning in it. All because in 2019 it invested in OpenAI, a competitor of the then much-known DeepMind. It has already proven to be one of their best investments. In total, they spent only $10 billion, and their main competitor lost $100 billion in capitalization because of it the other day, trying to imagine something at least very similar to the public.

Arms race

There is feverish excitement around ChatGPT, this AI has already found dozens of applications, from writing theses and conducting interviews to debugging programs and writing rap lyrics. It is widely believed that it will be able to take search engines to a new level. A chatbot is able to provide complex answers to questions by synthesizing information contained in billions of words pulled from the Internet and other sources. Working with a bot can give the feeling of a smoother, in some sense even cooperative search. This could potentially be a huge new niche.

But the way ChatGPT works is in some ways at odds with the idea of ​​a conventional search engine. It is supposed to accurately extract and transmit information found on the Internet. But, as it turned out, the bot easily generates new false information. Inventing something that really doesn’t exist anywhere. Its underlying algorithms do not extract data from the fact or reference base directly, but instead generate strings of words that statistically resemble those seen in its training data. The presence of truth is not taken into account anywhere. So, this search engine will be, to put it mildly, unreliable.

Despite this (huge) problem, several titans of web search, as well as a whole host of smaller companies, are already racing ahead. Microsoft is now trying to add the underlying technology to its Bing search engine and its Edge browser. In their view, this will allow them to finally gain an advantage over Alphabet with its perennially dominant Google and Chrome.

People are already enthusiastically waiting for it. In January, the ChatGPT chatbot had over 100 million active unique users! In the two months since its launch, the audience of the service has grown 100 times, which is the fastest growth in the history of the Internet (the previous record was TikTok, which took 9 months to reach 100 million users).

The other day, Microsoft announced that in addition to the usual search in Edge and Bing, there will now be a “Chat” option, which adds a chatbot interface to which you can ask questions. After this announcement, the popularity of the Bing mobile application increased 10 times, and the value of the company’s shares began to rise again, almost reaching $2 trillion. Even specialists from Google are now trying to find work in OpenAI and are leaving the search giant, in which the future seems less promising to them.

technical giants around the world are preparing their responses. Even China’s leading search engine Baidu says it is working on a Chinese-language bot similar to ChatGPT. Several startups have also launched their search engines with a robot-like chat interface. These include You.com, Perplexity AI and Neeva. But the main word here, of course, belongs to Alphabet, whose business seems to be threatened by the new mania.

Google’s answer

Alphabet certainly isn’t going to let Microsoft (or anyone else) take its crown as the #1 search engine in the world. And the company took the issue of II-chatbot development very seriously.

In December, Google issued a code red for the existence of ChatGPT. This means that management interprets it as a potential threat to its business, and the company must be ready to take decisive action. The situation is so serious that, according to the Times, Larry Page and Sergey Brin, who remain Alphabet’s controlling shareholders, came to the office and “gave advice” to the company’s leaders, approving their plans and offering new ideas during meetings with executives. As a result, Google announced in late January that it was laying off more than 12,000 employees to streamline costs and said it plans to focus on artificial intelligence as an area of ​​top priority.

Bard chatbot

The other day, Google announced that it is going to release its chatbot named Bard “in the coming weeks”. The company has been working on a sophisticated search AI called LaMDA for a long time: it was first announced back in May 2021. As reported, Google plans to release LaMDA in 2024, and already this year it can demonstrate up to 20 products that use this technology.

Sundar Pichai wrote in his blog on February 6 that Bard is now available to “trusted testers” and is designed to hide “the breadth of the world’s knowledge” behind a simple dialog interface. Bard uses a simplified version of LaMDA. As the CEO of Google says, this will make it possible to offer the chatbot to a larger number of users.

Notably, Pica hasn’t announced any plans to integrate Bard into the search box we’ve come to associate with Google, which provides the company with its main source of revenue. Instead, he demonstrated a new and very neat use of AI to synthesize new results. If the question is formulated in a characteristic form suitable for a chatbot, the user will receive generated text instead of the usual results. And if he doesn’t like it, he can scroll down to try to find something in the exiles. If there is no consensus on the correct answer, Bard will synthesize several options that reflect different points of view.

For example, soon when you type in Google “Which is easier, learning to play the piano or the guitar?”, at the top of the window you will see the lines:

Some say that playing the piano is easier because the movements of the fingers and hands are more natural. Others say that guitar chords are easier to learn, so you can start playing a simple tune within hours.

It usually takes 3-6 months of regular practice for guitar and 6-18 months for piano to reach a good level.

Pichai also said that Google is going to make Bard’s underlying technology available to developers through an API, just as OpenAI is doing with ChatGPT. He did not name the terms.

However, Bard’s public presentation turned out to be a downright failure. Alphabet lost $100 billion (!) of market value because of it. In a commercial, its new chatbot gave the wrong answer to a question about the James Webb telescope, which was spotted by Reuters on Wednesday (but unfortunately missed by Google’s advertisers). This disappointed the public and raised fears that the company was starting to lose out in competition with Microsoft.

As a result, Alphabet shares fell 9% in intraday trading. And in general, the company has already lost more than $500 billion, -30% of the value, in a year. While Microsoft remained almost the same (-10%), and since the presentation of ChatGPT has grown even more, so that it is now worth $750 billion more than Google. And once, less than three years ago, the two companies went hand in hand .

A $100 billion mistake

Where does this lag in the field of AI come from? Google, by its own admission, has chosen to tread very carefully when it comes to adding the technology behind LaMDA to finished products. These AI models are known to generate incorrect information, and are prone to exhibit racial and gender biases and repeat hate speech “learned” from the Internet. The company wanted to avoid scandals if it appeared, say, in a Google search.

These limitations were highlighted by Google scientists in a 2020 research paper, where they argued for caution with text generation technologies. Meanwhile, other developers who worked on LaMDA’s technology were frustrated by Google’s reluctance to implement their AI into products, and left the company to create startups that use similar algorithms.

In this sense, Microsoft was in a very good position: firstly, ChatGPT is a new separate product, and secondly, no one associates it with the existing brand. If AI was received very poorly, Microsoft’s core business would not suffer in any way.

The appearance of ChatGPT and the way it was received by the public must have inspired Google to accelerate the introduction of text generation technologies into its products. And the scientists who advocated for the creation of “ethical” AI, who issued this 2020 paper that delayed the release of LaMDA, were fired this year.

And here’s what ChatGPT himself thinks about his competitor

Unsolvable bot problems

Google’s lag behind Microsoft seems to be very significant. ChatGPT is already admired by everyone, while Bard gives wrong answers even in his presentation, and 100 million active users on the same place is unlikely to gain. But does this mean that Google’s era as the leader of search engines is going into the sunset, and Bing will be our new king?

Most likely not. And the gap of $750 billion in market value is clearly exaggerated. Yes, OpenAI definitely beat DeepMind. And the chatbot in Microsoft products will be better. But many experts say that such robots alone will never become good search engines, and will not replace people’s usual habitual search. All because AI chatbots have three problems:

  1. As mentioned earlier, they compile the information they find and in the process of generating the text often “make up” its fragments, bearing a full otsebyatya, even very convincingly. As a result, they cannot be relied upon as a source of information. And you can’t check where this or that data came from by following a link and checking the source. Therefore, if the robot made a mistake, it will be possible to check the facts only through the good old search engine.

  2. II-chats are very difficult to update. Basically, if new data is released, the entire model needs to be retrained so that it learns to correctly incorporate it into its neural networks. Retraining such models on Google-scale data would cost billions, and would have to be done regularly, otherwise the data would become outdated compared to search engine data. In addition, the active work of a searcher on ChatGPT will cost ten times more than the Google search engine, because he will need to generate each new answer. It is not clear how this will pay off.

  3. Chat interfaces are incompatible with the current advertising revenue model of search engines. Currently, Google and Bing top ads are shown based on the given search queries, as well as links to promoted sites. How do boots fit into all this? If you need the audience to follow links to sites to get relevant information, a text generator will be rather redundant. It may take new forms of advertising to make chat-style search interfaces viable, but it’s not yet entirely clear what those might be.

Therefore, most likely, rumors about the rapid fall of Google are exaggerated. Chatbots will not capture the entire market of search engines. But if the company gains a reputation as a “lagging behind” in terms of technology, it can have a cascading effect. Top employees are already leaving Google and DeepMind, which it owns, and others may follow. Without a reputation, the most progressive and most developed money from investors cannot be obtained either. Tesla shares have soared for several years on this reputation alone. If the public believes that Bing and ChatGPT are the future, that will be enough in itself. Google must now do everything to prevent this from happening.

Experts also warn that tools like ChatGPT can create a lot of problems for search companies by flooding the Internet with spam texts created by artificial intelligence and optimized for search engines. It may take some time to figure out how to prevent all the issues from being filled with nonsense generated by language models. Finding a way to curate good information to keep search relevant will require new approaches from Google and Bing. How long it will take for methods to combat text AI spam to be invented and tested is unclear.


PS Thousands of cool vacancies in the Russian Federation and abroad – in the telegram bot getmatch. You set the desired salary and the best offers come to you, and our employees help you pass the interview. No resume or portfolio required, setup takes less than 30 seconds. And it’s also a great way to monitor the labor market.

What else to read

If you look events in Mykolaiv – https://city-afisha.com/afisha/

Related posts