🍿AI highlights from this week (2/10/23)
The BigTech AI battle begins, AI replaces designers, Chatbot for everyone and more...
Hi readers,
Here are my highlights from the last week in AI, when Microsoft declared war on Google’s search business with the help of OpenAI.
P.S. Don’t forget to hit subscribe if you’re new to AI and want to learn more about the space.
The Best
1/ Search Wars: Attack of the Chatbots
This week was a big week in AI for Big Tech with Microsoft and Google both announcing product updates related to AI. First up was Google on Monday with a blog post to pre-announce their new chat product Bard:
The post might have been an attempt to steal Microsoft’s thunder who planned a last minute press event on Tuesday to announce their OpenAI integration. In the post Google’s CEO Sundar Pichai was quick to remind everyone that we wouldn’t be where we are today without Google’s contributions to AI:
At the same time, advanced generative AI and large language models are capturing the imaginations of people around the world. In fact, our Transformer research project and our field-defining paper in 2017, as well as our important advances in diffusion models, are now the basis of many of the generative AI applications you're starting to see today.
Pichai then went on to share a few details of Bard, a new chatbot Google has developed including that it is powered by their large-scale language model1 LaMDA. Bard seems to be Google's answer to ChatGPT but with only "trusted testers" given access to the product, we all had to eagerly await a demo from Google at an announcement they scheduled for Wednesday.
On Tuesday, Microsoft quickly followed with their own blog post and press event to announce several new AI features across their Bing search engine and Edge internet browser.
Microsoft’s AI integrations are powered by the latest OpenAI model, which is more advanced than ChatGPT. The new features include an AI integration in Bing that allows customers to ask questions and get answers from a chatbot:
And an AI integration in Microsoft’s browser Edge that summarizes webpages and helps you write content for social media posts:
This launch was a clear indicator of Microsoft’s willingness to take risks and go all-in on AI, challenging Google’s monopoly in search for the first time in decades.
On Wednesday, it was Google’s turn to mount its defense with a much anticipated AI press event held in Paris. Surprisingly Google only spent 4 minutes of the 40 minute event talking about Bard, which still isn’t available to the public. The press event failed to impress and hit a few hitches along the way, including a VP of product who couldn’t find the phone she was supposed to demo with!
After the disappointing event, things only got worse for Google when reporters discovered that one of the demo videos of Bard incorrectly answered a question about the NASA’s James Webb Space Telescope.
Given how much Google has been talking about the importance of AI providing accurate answers to users, it was ironic to say the least that they would make such a public fumble. The day cost Google $100B in market capitalization with their stock trading lower for the rest of the week.
What was more costly was the damage Bard’s fumble and the haphazard event did to Google’s reputation as an AI leader, which is undoubtedly coming into question now. While Sundar Pichai has been a tremendously successful peace time CEO, it seems he is struggling to lead Google into the impending war.
Round 1 of the Big Tech AI War goes to Microsoft.
2/ Is AI going to replace designers next?
This week a former Google AI researcher and a Facebook designer launched Gailieo, a product that can generate user interface designs just by describing them using generative AI.
I haven’t had a chance to try the product out myself yet, but as a product manager who often relies on designers to bring my ideas to life, I’m excited for what Galileo promises to do. Imagine being able come up with an idea and, without having any creative abilities, quickly create a mock-up by just describing the screens you want to see. For designers, the idea is that Galileo will also free up their time to focus on deeper more complex user experience problems.
At the same time, I can understand why some designers might have mixed feelings about Galileo…
Maybe product management will be next?
3/ ChatGPT as a service
Microsoft announced that it would allow companies to create their own ChatGPT, trained on their internal data, which means you may soon be seeing Chatbots greeting you at at many of the online services you use every day.
Microsofts aggressive pace of innovation in AI has got me wondering about where the defensible moats are for a startup building in this space. Microsoft has both the tight integration with OpenAI and the most significant enterprise distribution of any company. It has also been known to clone successful products to take advantage of this distribution for example, Microsoft Teams, a clone of Slack.
It turns out that creating a Chatbot for your own company’s data is not as hard as one might assume. This week Quora launched its own Chatbot that uses Open AI’s GPT as one of the underlying models users can choose from:
In fact, with some sneaky prompt engineering, people have already reverse-engineered exactly how Bing’s new chatbot, which OpenAI’s model powers, is conditioned to respond to users:
If you’re interested in learning more about how to make your own Chatbot, check out this great post from Lenny’s newsletter on how to make a chatbot for your newsletter:
4/ Is Google’s core business under threat?
Google is in the business of making money from you clicking on sponsored search results that take you to other websites who pay for those clicks. Chatbots like ChatGPT, on the other hand, give you the answer to your question directly without you ever having to visit another website. These two approaches are fundamentally at odds with each other. If Google replaced their search bar with a chatbot, their core revenue stream would go to zero. This is why chat-based search poses such a threat to Google.
This week in my favorite podcasts, “All In”, the “besties” had a riveting discussion about what AI means for Google search business. I highly recommend listening to the podcast from 33 mins onwards if you’re interested in how AI might impact the search and publishing business.
One of the perspectives I found fascinating was when Chamath Paliyahiptiya suggested that Google should license exclusive deals with publishers to use their content in their AI answers. In this world,9 the same publishers would also prevent other large language models like Open AI from crawling the same content to train their models, which of course would be bad for Microsoft’s Bing. It would be a shrewd move by Google and not that dissimilar from their exclusive multi-billion dollar deal with Apple to be the default search engine on their devices. Publishers may also have no choice but to go with the devil they know rather than being completely disintermediated by a AI chatbots that don’t send any traffic their way. At the same time, it would be an unfortunate outcome if Google, the company’s whose mission is to organize the world information, used their monopoly in search to prevent competitive innovation in AI.
That is unless we come up with an innovative new revenue model that allows AI to give us the answers we need efficiently while still rewarding publishers. To do that we will need to solve attribution of AI answers to the websites that the language model used to come up with those answers. Neeva, a search engine founded by a former Google executive has made progress in this direction. Their AI powered answers to searches include references to relevant websites:
The challenge however is that search engines like Neeva or Bing have much less to lose experimenting with these approaches than Google does. Even a 5% loss in market share to rivals would be hugely impactful to Google’s bottom line as the besties pointed out in their podcast.
So is Google’s core business under threat?
Six months ago, no one would have guessed that tech influencers would be switching to Bing for search or installing Edge for browsing. While that may be a drop in the ocean of search queries going to Google, it surely can’t be a good sign.
I thought this quote from Satya Nadella’s interview with the Financial Times best summed it up:
“There is such margin in search, which for us is incremental. For Google it’s not, they have to defend it all”
The Rest…
A few other updates in the world of AI from this week:
- The latest version of OpenAI’s GPT displays the ability to “think” as well as a nine year-old.
- Google shares another AI research update:
- Thoughts on why Google and Microsoft’s cloud divisions are incentivized to invest in AI companies
- Noteworthy ideas from an influential AI researcher
- Open-source ChatGPT competitor:
Tranformers2 reading list:
David Guetta uses AI to get Eminem to rap on his latest EDM track
Finally, In case you missed it, my interview with Ammaar Reshi on how he published a children’s book in AI in a weekend:
That’s all for this week!
Thanks for reading The Hitchhikers Guide to AI! Subscribe for free to receive new posts and support my work.
A large-scale language model (LLM) is a type of deep learning model that is trained on a large dataset of text (e.g. all of the internet). LLMs predict the next sequence of text as output based on the text that they are given as input. They are used for a wide variety of tasks, such as language translation, text summarization, and generating conversational text. Open AI’s GPT-3 (General Pre-trained Transformer 3), the language model that powers Chat-GPT, is an example of a generative LLM that uses the Transformer architecture, enabling it to be trained on a massive text dataset of hundreds of gigabytes using 175 Billion parameters (weights assignments). You can learn more about LLMs in my deep dive into deep learning part 3. ↩
In 2017 Google’s AI researchers published a paper “Attention is All You Need” where they proposed a new neural network architecture call the Transformer. This architecture created a step change progress in AI which enabled Large-scale Language Models like ChatGPT and Generative AI like Stable Diffusion. You can learn more about transformers in my part 3. of my series on the origins of deep learning. ↩