OpenAI freed🎊, Copilot in your IDE💻, Bill’s AI exuberance☀️
You and your AI are now free to move about the internet
Hi there! Welcome to our 42 wonderful new readers since the last issue. 👋 Let’s catch up on some AI news!
1. OpenAI now supports plug-ins; allowing ChatGPT to broad access to internet
(3 min read) OpenAI is rolling out plug-in support for ChatGPT, starting with Expedia, OpenTable, Kayak, Klarna Shopping, and Zapier. This will allow ChatGPT to access third party data on the internet. Eventually, this can turn ChatGPT into an interface for a wide range of services and sites. See the launch announcement here and join the developer waitlist here.
Thoughts: This feels very much like the launch of the App Store. When announced, there were 100 apps on Apple’s new ecosystem. Today, there are millions, with developers earning $60B in 2022 alone.
2. Google slowly rolls out Bard public test
(2 min read) If you signed up for the waitlist, you should get access to try out Bard, Google’s answer to ChatGPT. Google is proceeding slowly (US and UK only), given it has more to lose than Microsoft and remembering its $100B announcement gaffe.
Thoughts: The (beta) launch is underwhelming. See this Bard vs. Bing vs. ChatGPT match-up from the Verge (18 min read) and a few of my thoughts on Google’s moats further in this post.
Meanwhile, I asked Bard to help me code:
ChatGPT seems to be many steps ahead of Bard: capabilities, speed of development and iteration, public and developer interest, developer access, focus, partnerships, and more... but the AI wars have only just begun.
3. GitHub Copilot gets a new ChatGPT-like assistant to help developers write and fix code
(5 min read) GitHub is using OpenAI’s latest GPT-4 model to go way beyond auto-completing comments and code. Copilot X gets chat and voice support. Copilot can analyze the code for security vulnerabilities or explain how blocks of code work, all from within your IDE.
Thoughts: People have already been co-creating code easily with ChatGPT. This will become even more streamlined, allowing millions of new “no-code” programmers to begin creating and iterating software.
4. Bill Gates says OpenAI’s GPT is the most important advance in technology since 1980
Microsoft founder Bill Gates says that OpenAI’s GPT AI model is the most revolutionary advance in technology since he first saw a modern graphical desktop in 1980. Read the complete post here (20 min read, but it’s worth it).
“The development of AI is as fundamental as the creation of the microprocessor, the personal computer, the Internet, and the mobile phone. It will change the way people work, learn, travel, get health care, and communicate with each other. Entire industries will reorient around it. Businesses will distinguish themselves by how well they use it”
- Bill Gates
Thoughts: Just going to add that Bill casually skipped over the internet and mobile phones and went straight from GUI to GPT. Whoa. Tell us how you really feel, Bill.
Oh, and I’m inclined to agree.
5. Microsoft adds Image Creator to Bing, plus GPT-4 now available in Azure OpenAI Service
(5 min read) Bing Image Creator, available in the new Bing preview on desktop and mobile, as well as in Edge, is powered by an “advanced” version of OpenAI’s DALL-E model. Image Creator can generate an image based on a description and context from the user.
Quick hits
Microsoft Now Claims GPT-4 Shows 'Sparks' of General Intelligence
GPT-4 Hired Unwitting TaskRabbit Worker By Pretending to Be 'Vision-Impaired' Human
A History of Generative AI: From GAN to GPT-4 (long read)
Today’s Deeper dive (Bard vs. Bing)
Although Google and Bard seem to be falling significantly behind Microsoft and GPT, there are a few moats that Google can leverage as the AI wars continue:
1. The way we search: We have adapted “Google-speak” as a way to prompt the search engine, smushing words in a semi-neutral tone: “bad traveling love death and robots”, “apple M1 chip intel or apple”, “tmobile outage map”, “pinnacles national park bats”, “stratechery bing” (that brings you largely up to speed on my searches*). Although the new Bing can handle these somewhat non sequitur prompts, it is trained on natural human language and questions, and could perform suboptimally outside those conditions.
2. Demand and supply switching costs: Getting people to change the way they search adds cognitive load to the “demand” side costs of switching. And on the supply side, Google is also quite magical (and often superior to ChatGPT/Bing) at answering machine-like queries, and at lightning speeds.
3. Instant gratification: The modern day user doesn’t like waiting. Over ten years ago, Google shared that “a 400ms delay leads to a 0.44 percent drop in search volume” (Urs Hoelzle, in 2012 Google blog). I doubt that we collectively became more patient in the last decade. Google indexes the internet whereas a chatbot responds, word by word, “completing” your specific request. Because of this difference in methodology, it may be difficult to get ChatGPT and Bing to generate a response within the ~0.4 second waiting time we’re used to with Google.
4. One is better than two: We do already pose some of our searches as questions. I googled “when was BASIC created” to confirm our timelines in a past issue. And (at least for now) it does feel like an AI-powered search can better answer these more complex questions. So people may be drawn to chatting with their search engine for more open-ended questions but still expect instant answers to many simpler, navigational or informational queries. Right now, that means Bing/ChatGPT for the former and Google for the latter.
But it is hard to imagine that people will settle into a pattern of performing informational searches in one place and have in-depth discussions with a sophisticated chatbot in another if both are available in one place. Google can deliver the chatbot eventually, but Microsoft is unlikely to deliver a superior informational and navigational search experience.
5. Ecosystem: Whether we like it or not, we live in Google’s world. The daily use of Google services firmly entrenches many millions of users within G’s ecosystem. YouTube, Gmail, Maps, Chrome, Android, Home, Drive/Docs could all be retrofitted with AI and used time and time again to bring (even momentarily) lapsed Google Search users back. Of course, Microsoft has quite the ecosystem itself -- and we can cover it in the next issue...
*Interested in understanding the share of your Internet search queries that fall into navigational, informational or open-ended? Check your search history through Google’s My Activity; filter by “Search”.