The Implications of Today's HUGE AI Announcements
GPT-4; Generative AI for Gmail, Docs and Slides -- This is one of AI's biggest days Ever
Did all the A.I. companies come together in secret and decide to announce ALL of their new products on the same day?
Were they trying to get news in before another financial crisis takes over the news cycle?
In case you missed it: Google announced Generative AI for Google Workspace (formerly G Suite) — Gmail, Slides, Docs, Sheets, Images, ALL of it. Not to be outdone, a few hours later, Open AI announced GPT-4 — perhaps the most advanced large language learning model ever.
Now that I’ve had some time to digest the news, I want to break down the two big announcements and, more importantly, what they mean for consumers AND for the future of tech and A.I.
Let’s start with Google, then move on to GPT-4.
Gmail, Slides, Docs, Sheets, Images Just Got 10x Smarter
If you or your company pay for Google Workspace (any site with a custom domain/email with Google), you will soon get Generative AI for Gmail, Docs, Sheets, Slides, Images -- EVERYTHING. If you haven’t watched their video announcement yet, stop reading and watch it first. (I included it right above this text.)
No word on normal Google users (anyone with a gmail.com email address), but I suspect they may get these features down the line.
Here are just a few of the MANY things you will be able to do with their new AI:
Draft, reply and summarize your emails. Literally you can say, “summarize this email conversation” to Gmail, and it will summarize it.
Brainstorm, proofread, write, and rewrite in Google Docs. It’s interconnected with your Gmail, so you could ask Google to write a brief about an email chain and it’ll place that summary in Google Docs. You can highlight a section and ask Google to rewrite it. You can even give it instructions on how to sound.
Auto-generated images, audio, and video in Slides. It will build a presentation off a Google Doc. Imagine writing out the bullet points for a fundraising deck and having Google just… build it. Including the images!
Now stop imagining it, because IT IS REAL.
Tome — the super cool AI that can generate presentations and just raised $43m — has a competitor now. It will be fascinating to see these two compete.
Autocomplete and create content within Sheets. Go from raw data to insights and analysis via auto completion, formula generation, and contextual categorization in Sheets. Imagine writing 1000 personalized emails in a single spreadsheet in a few minutes, or asking AI to pull out key numbers or context from a messy Google Sheet.
Oh yeah, you don’t have to imagine it anymore!
This isn’t all — it can draft notes during Google Meets, help you write messages in Chat, and a whole host of other things we likely haven’t even seen yet. It’s the biggest upgrade to Google’s productivity suite since… perhaps ever.
Google’s new AI features will be available to select testers this month, with a fuller rollout to Google Workspace customers shortly thereafter. If you aren’t using Google Workspace yet, I’d sign up now. It starts at $6/m. It’s a steal — I’d personally pay more than that just for the AI alone.
It’s also not the only announcement Google made today — they also announced AI capabilities in Google Cloud and a Generative AI App Builder for developers. I won’t focus on those announcements today, as I suspect Google will have even more developer announcements coming soon.
Thoughts: Productivity Will Never Be the Same
A few takeaways from Google’s announcements:
The idea that you had to build a presentation from scratch or write any document from scratch is dead. Or it will soon be. Google Workspace is used by hundreds of millions of people (maybe more?), and generative AI being prominent in all of their apps will supercharge adoption.
The cross-platform nature of Google’s new AI is perhaps its biggest benefit. It allows you to no longer copy and paste content from one app to another, and instead think of them as all one application. Taking content from an email and having AI rewrite it in a Google Doc? Spinning up a presentation from a Google Doc? *Chef’s kiss*
The cross-pollination of AI tasks is a huge benefit for Google.
Should you start questioning if the emails you’re receiving were written by AI? Yeah, you probably should. Almost every email is going to have at least some amount of AI in it now.
Microsoft will not sit idly by. They have an AI productivity announcement event on Thursday, where I expect they will announce similar features for products like Microsoft Word. The AI wars are heating up even more. AI innovation will be the ultimate winner.
GPT-4: Way Smarter and Available Now in ChatGPT Plus
It’s real. GPT-4, the next version of Open AI’s large language model has been announced. And you can play with it now on ChatGPT Plus.
You can go here to read Open AI’s announcement post. I also recorded a TikTok breaking down the announcement. But in summary:
GPT-4 is a SIGNIFICANT upgrade to GPT-3.5. It can, for example, outperform GPT-3.5 (the tech that was being used in ChatGPT until today) on almost every type of exam. For example, while GPT-3.5 scores in the bottom 10% of test takers for the bar exam, GPT-4 scores in the top 10%.
It’s not just the bar exam, either — it can do AP Calculus, AP Chemistry, and AP Physics much better now. Here’s a graph of GPT-4’s performance vs. GPT 3.5:
This one is my favorite. GPT-4 supports VISUAL inputs. What does that mean? Well, you can literally upload an image and ask GPT-4 what it’s about, and it will tell you. 🤯
This is example that blew my mind:
There’s a lot of tech that goes into understanding the context of an image. I sat on the board of an AI data training company for 3 years, so I have a deep understanding of the work that had to go into making GPT-4 be able to “see”. It’s shockingly impressive and I can’t wait to find use cases for this in my daily life.
GPT-4 has more steerability, giving you the ability to give your text more personality instead of the more generalized tone that ChatGPT gives you. This matters because ChatGPT can often sound “generic”. GPT-4 almost completely eliminates that problem.
Existing AI detector tools fail when trying to detect GPT-4. I literally just tested this — two different AI detectors thought some GPT-4 text I generated was 100% human.
GPT-4 can take in 25,000 words of context — 8x more than ChatGPT. You can analyze larger volumes of content or write longer prompts with greater context.
GPT-4 is also more accurate, though nowhere near perfect. It hallucinates (gets facts wrong) 40% less than GPT-3.5. It’s still getting things wrong one out of five times though, so remember that before using GPT-4 as a source of truth:
GPT-4 is available now in ChatGPT Plus with a usage cap (it’ll be constrained at the beginning, but will grow). Access to the API for GPT-4 is waitlisted, so you need to get on the waitlist to get access.
Thoughts: Expect GPT-4 to be EVERYWHERE
Do you remember the world before ChatGPT? Because I barely do. No joke: I think the trajectory of our society itself was shifted by ChatGPT. GPT-4 just confirms it.
GPT-4 is more sophisticated and nuanced than its predecessor. In some initial testing, the difference in output is stark. I asked GPT-3.5 and GPT-4 to write me a unique essay about the history of Super Mario. GPT-3.5 wrote a good but slightly generic essay. GPT-4 came up with a title, subtitles, and its own tone of voice. It wrote for longer and made the best use of it the restriction I gave it (limit itself to 5 paragraphs).
I posted the full essays GPT 3.5 and GPT 4 wrote on my Twitter (to save space here). But seriously, it’s night and day between these two essays. GPT-4 is the real deal:
Open AI did a good job tempering expectations during its announcement — it’s not perfect, and after the initial shock of its strength, you start finding its flaws. But if this is how quickly AI technology advances… how long until we get programs that are 10x better than humans at most major tasks?
I feel a deep sense of awe and a minor amount of dread when I look at GPT-4 — and the rapid advance of AI in general. If you have some mixed emotions about the rapid advance of AI (like me), that means you’re human.
I’m pushing hard on the AI side, as are the Octane AI team. Stay tuned — I have a few announcements (both business and personal!) dropping soon, as does Octane AI. I will share them all here on The Social Analyst.
In the meantime, I’m always here to talk about the latest in AI.
Can't help but wonder... when Apple?
When will it be able to write something like this, a hilarious short story of Mario turning 40. https://www.newyorker.com/magazine/2022/04/25/mario-not-so-super-at-forty