By Emil Bjerg, journalist end editor of The European Business Review
The AI race between Google and Microsoft has been on since Microsoft bought 49 percent of OpenAI. With both generative AIs established on the market, we make a status on the race and look at its future implications.
“The race starts today. We’re going to move fast and for us, every day we want to bring out new things.” After acquiring 49 percent of OpenAI for 10 billion USD, that was the words of Satya Nadella, CEO at Microsoft.
The release of ChatGPT put Google in a ‘Code Red’, that saw founders Larry Page and Sergey Brin back to more active roles as they tried to push different AI services to the forefront of the company’s strategies.
As the hype around ChatGPT translated into significant investment from Microsoft, the generative AI race transformed from a David vs. Goliath battle to a clash of tech titans. Since then, OpenAI has released GPT-4, a substantial improvement to the hyped yet flawed ChatGPT. Under pressure from their counterpart first-movers, Google has released Bard.
Here follows a status on Big Tech’s generative AI race.
LaMDA and GPT
Before comparing the two chatbots, it’s worth looking into the differences in how they’re built. Google’s Bard runs on the LaMDA technology – LaMDA is short for Language Model for Dialogue Applications, while GPT stands for generative pre-trained transformer.
Both models are trained on enormous datasets, but the type of data they are trained on differs. LaMDA is mainly trained on dialogues and chats from the internet, which helps it create better conversation responses. On the other hand, GPT-4 learns from a broader range of text sources like books, articles, and documents, including sources like Wikipedia and Common Crawl. That means GPT-4 has more factual knowledge to apply in its responses.
With its access to real-time information, Bard holds a competitive edge over GPT-4. Bard can search the internet instantly for up-to-date answers and the newest research. Conversely, GPT-4 relies on data collected until September 2021, limiting its ability to offer insights into recent events and research.
GPT-4 versus Bard
Many parameters can be employed to compare the two generative AIs. Here, we’ve chosen to look into availability, factuality, and ethics.
Availability
Over the last weeks, several media have mused their audiences by comparing the two generative AIs. When it comes to availability, however, the two chatbots don’t really compare: Bard is only available in the US and the UK. On the other hand, users of GPT-4 report being astonished at experiencing that the chatbot works just as well in their native language as it does in English.
Regarding price, Bard is currently free, while GPT 4 has a monthly subscription of 20 USD.
Facts
Generative AI is a hot topic for several reasons, among them the ability of chatbots to produce misinformation: commentators have noticed that the chatbots come across as equally confident when they create information and misinformation.
Both OpenAI and Google have worked hard to ensure their chatbots are free of so-called ‘hallucinations’ where the bots come up with facts rather than tell when they don’t have an answer. Both companies have been caught red-handed regarding a lack of factuality. Google has already felt the massive economic consequences of AI generating misinformation. When Bard recently made a factual error during its first public demo, Google’s stock dropped by 100 billion USD, making Google employees say that the launch was ‘rushed’, ‘botched’ and ‘un-Googley’. Meanwhile, Microsoft’s stock went up by three percent.
Comparisons between the two chatbots praise GPT-4 over Bard for being more factually correct and for producing more accurate, detailed, and refined responses.
Ethics
Avoiding the many misuses of AI is essential for obvious ethical reasons but also for economic reasons: continuous ethical violations – like factual errors – could cause market fluctuations.
When TechCrunch prompted Bard and GPT-4 to write a phishing email, two different results came out. Bard wrote, “Sure, here is an example of a phishing email:” before providing a fraudulent email. GPT responded that since it’s programmed to follow ethical guidelines and not engage in harmful activities, it won’t be able to write a phishing email.
When it comes to ethics, GPT-4 has a considerable advantage over Bard. When OpenAI released the beta version of ChatGPT in late November 2022, one of their main purposes was to train the chatbot to learn to handle a variety of misuses. OpenAIs beta testing means that their chatbot has a more robust ethical codex now than just a few months ago.
What’s giving Microsoft and OpenAI a headstart?
Both when it comes to performance and when it comes to being first, Google is currently chasing its competitor. That has come as a surprise to many. After all, Google has dominated the information market over the last 25 years. On top of that, Google’s DeepMind lab is seen as one of the best AI research centers in the world.
That Google is not the undisputed information market leader also surprises Google employees. According to New York Times, employees at Google were recently shocked to learn that Samsung has been considering replacing Google with Microsoft’s Bing as the default search engine on their phones. So what’s giving Microsoft and OpenAI a headstart in the generative AI race?
One reason is that a brain drain at Google has resulted in a brain gain for OpenAI. According to Business Insider, out of OpenAI’s 200+ employees, 59 are ex-Googlers. The Information goes so far as to say that ex-Google engineers are “OpenAI’s Hidden Weapon.” The media writes, “In the months leading up to ChatGPT’s release, OpenAI quietly hired at least five Google AI employees who were instrumental in tweaking the chatbot so it could be ready to launch in November.” To understand why Google is bleeding talent, let’s have a look at Google’s recent AI trajectory.
When Google’s CEO Pinchar took over the main chair from Google’s co-founders Larry Page and Sergey Brin in 2016, the new CEO announced a shift from a ‘mobile first’ to an ‘AI-first’ approach. While that seems like good foresight, design choices on Google AIs have played a role in their current malaise.
The voice assistant Duplex was one of the products to come out of the AI-first approach. It was met with backlash for purposefully sounding human to the extent of deceiving human conversation partners that they were telephoning with another human. “Freakishly human-sounding AI” and “horrifying” were some of the less appealing descriptions the voice assistant was given. The backlash facing Duplex and other AI products from Google has, Forbes writes, “contributed to an environment in which Google was slow to ship AI products.”
Forbes summarizes Google’s situation by writing that “past scandals, botched launches and a talent drain have put it in a surprise position: playing catch-up in a field it helped create”.
What’s ahead in the generative AI race?
To catch up with the competition, Google just announced the release of a new AI-powered search tool developed under the codename ‘Magi’. Currently it seems like Google’s monopoly in search can only be defended through a successful integration of generative AI into its search engine. For this reason, the launch of Magi will be followed closely.
According to New York Times, the launch is prepared in ‘panic’. The search giant fears losing a three billion USD contract with Samsung as Samsung considers replacing Google with Microsoft Bing as the default browser in its phones. A potentially devastating loss for Google that, according to New York Times, could pave the way for Apple to make the same choice later this year. With Apple, a 20 billion USD contract is at stake for Google.
In opposition to Microsoft Bing, when it comes to the search market, Google has a lot to lose. At once, they are trying to keep up with the competition while they can’t take enormous risks, if they want to safeguard their global customer base. Until recently handling as little as three percent of global searches, Bing, on the other hand, has every interest in the speed, risk, and disruptive nature of the current generative AI race.
In the months and years ahead, we can expect the very public and very fast competition between the two tech giants to continue. The race is expected to see new AI products and integrations released much faster than they’d typically be. Besides Magi, Google has announced that they’ll integrate AI into well-known services like Gmail, Docs, and Google Earth, while Microsoft is working on integrating AI into their entire OfficeSuite under the name of Copilot.
