Google announces something, but OpenAI steals the thunder with GPT-4

Greetings!

Here is the latest reporting from the trenches of the ongoing AI war…

I try not to report too much on the latest news, but I have no choice this week.

Google announces something big on Tuesday, and for a few hours, they are the talk of the town…

Then OpenAI launched GTP-4 a few hours later, and Google lost its mojo.

It is hard to describe in words how my head feels right now…

Today’s headlines:

• Google announces something, but OpenAI steals the thunder…

• What is GPT-4 anyways?

• The TL;DR of GPT-4

• Google announces something very interesting

• What are some implications of Google’s new AI plans?

• My mind is blowing my mind

• The robots are coming to football (soccer)

• Quick bites - interesting AI news

• Shameless plug - my new newsletter

Get yourself a pint of coffee, and let’s dig in!

Google announces something, but OpenAI steals the thunder…

Google actually announced something cool this time.

But OpenAI/Microsoft clearly has a better marketing and PR team and steals Google’s thunder (again).

Since the online media and the 269 other newsletters on AI have already done extensive reporting of the GPT-4 announcement, I will do a more limited reporting on the announcement and give some focus to Google as well.

In my opinion, the Google announcement is more interesting from a “news” perspective.

GPT-4 is, after all, “only” an iteration, while Google announces something new new!

What is GPT-4 anyways?

To clarify what GPT-4 is or means and how it all relates to ChatGPT.

OpenAI, the creators of ChatGPT, have several advanced AIs or Language Learning Models (LLMs).

GPT-4 is the latest iteration.

It is now bigger and better with more features.

ChatGPT is the “consumer-facing” tool we can use to access and use the AI (the GPT part).

The technical folks among us can access the GPTs using an API - an “interface” between the brain/AI and whatever program or tool they are developing.

The current ChatGPT, the free version, runs an advanced version of the “old” GPT-3 version, referred to as GPT 3,5.

GPT-4 is the “new” brain that ChatGPT eventually will run on.

It was also confirmed Tuesday that Bing’s Chat (Bing’s “ChatGPT”) uses GPT-4. This may explain the meltdowns, the “sentient-like” responses, the expression of love, etc.

GPT-4 is currently only available to ChatGPT Plus users (paying customers).

And even paying customers only have access to a limited number of functionalities and not the full range of promised features.

The TL;DR of GPT-4

1/
GPT-4 is a multimodal model, meaning you can prompt it with both text and images (it can “see”). You can ask ChatGPT questions like “what happens in this picture” or “what food can I make with the ingredients in this picture” or “what do the stats in this picture tell me.” OpenAI runs an exciting test project with the Danish company “Be My Eyes” helping the blind or those with low vision. Super interesting and something I will cover more in-depth in a future newsletter.

2/
GPT-4 is way way smarter than GPT-3. It can understand and solve more complex problems. OpenAI ran a wide range of tests where the AI had to go through various exams in multiple languages and basically aced them all. What tests/exams, you ask? How about passing the uniform Bar exam (lawyers) or LSAT (Law School Admission Test)? Compared to GPT-3.5, GPT-4 seems to pass many exams in the top percentile (link to the tests and more).

3/
GPT has a better chain of thought capabilities and, thus, can make killer dad jokes:

I found this on Hazel Lim’s Twitter feed (she also writes an awesome newsletter called What The Hype).

4/
GPT-4 has increased its risk mitigation and is safer. I have previously written about ChatGPT going woke. The models are only as good as their training and the “moderators” political leanings. Hopefully, this does not turn into another “Twitter nightmare.” Either way, more safety is a wise decision with such a powerful tool.

5/
GPT-4 has increased steerability, meaning you can control what outputs should be and the tone or personality of the answers. This is important, as it allows tool developers to set restrictions on what output they want inside their tool with integrated GPT-4. For example, if a tool integrates a Socratic method (answering questions with questions to help learn and think for yourself), this can be set either in the prompt or at a system level.

This will allow some serious prompt engineering magic.

6/
GPT-4 will be “connected” to the internet (currently not), meaning you can ask it to summarize an article by providing the link, and ChatGPT will work its magic. In the example given, they use a Wikipedia article. Maybe it is limited to Wikipedia, similar to what DuckDuckGo just released? (I spoke about DuckDuckGo in last week’s newsletter).

I don’t know if ChatGPT will be fully online-enabled and read/learn new things from the internet as they happen (I doubt it - it would lead to losing control over the tool very fast).

7/
GPT-4 is capable of handling over 25,000 words of text. That is a huge increase from before (from about 2000-5000 words). This will allow use cases like long-form content creation, extended conversation, document analysis and summary, and more. This, in my opinion, is huge! Combine this with the increased “brain” and “chain of thought” capabilities of GPT-4, and we have ourselves a “super tool.” I know several use cases, and I have a fun experiment I plan to try out in the near future using GPT-4.

8/
The most MINDBLOWING thing I saw GPT-4 can do - and this is crazy crazy stuff - is that it can create a webpage from a sketch on paper.

Say what?

In the live presentation, they showed how the presenter created a functioning webpage by simply giving GPT-4 an image of a VERY poorly sketched webpage and asking GPT-4 to create that page.

As a developer, a digital marketeer, and the owner of several websites, I already have 4-5 use cases ready once this capability is live!

As cool as this is, it also means that regular websites will be decimated by clever marketeers with a pro license of ChatGPT…

You can watch how they create the webpage in the live presentation video on Youtube, or you can watch Matt Wolfe of Futuretools.io similarly create a website by simply writing a one-line prompt into ChatGPT with GPT-4 (since the image function is not available yet).

Here are the various links to OpenAI’s information and presentation:

The GPT-4 product page.
Here is the full research article (with the stats, tests, etc.).
The live presentation of GPT-4 on Youtube.

Google announces something very interesting

I get that ChatGPT is the flavor of the week, the flavor of the year, and the flavor of the century.

And we can all see the power of Microsoft’s PR team.

But on Tuesday, Google announced that generative AI features would be integrated into Google Workspace (equal to MS Office).

This is actually a big deal!

In my media business, we only use Google tools. Docs, Sheets, Slides, Forms, Drive, Calendar, Meet, and Gmail.

And we are far from the only one!

This newsletter was written in Google Docs.

Microsoft is the big brother in the enterprise market with Word, Excel, Powerpoint, Teams, Skype, LinkedIn, and more.

But Google has over 3 billion users of their free tools (when you add Gmail for private persons into that count).

Let’s have a look at what is coming to Google Workspace!

• Draft, reply, summarize, and prioritize your Gmail

• Generative features (like ChatGPT) inside your Docs allow you to proofread, write, rewrite, etc.

• Analysis, formula generation, contextual categorization, insights, and more inside Sheets (its alternative to Excel)

• Generate images, audio, and video inside Slides (Powerpoint equivalent)

• Capture notes inside Meet (its Zoom/Teams alternative)

I must admit, looking at the video below, I get excited! Sure, it is polished and for promotional purposes. But as a Google Workspace user, this excited me more than Microsoft’s efforts. I am sure the AI features in MS Office, and the MS tools will be equally impressive.

What are some implications of Google’s new AI plans?

Although I am focusing on the implications of introducing generative AI into Google Workspace here, the same implications imply when Microsoft adds GPT-4 into its suite of tools.

My immediate reaction watching the video above was, “oh shit, there will be a blood bath among all the small stand-alone tools built with AI features.”

What does that mean specifically?

One of my favorite tools is Grammarly. I am not a native English speaker and Grammarly is essential to help me write this newsletter. It is 100% my words (not generative AI is used).

Grammarly helps me correct spelling errors and sometimes offers to change my sentences when they get too complicated or lengthy (it even changed this sentence 😃 ).

Grammarly has AI integrated.

I write these newsletters in Google Docs.

I have Grammarly integrated into Google Docs.

I pay for Grammarly Pro.

Will I need Grammarly once Google has introduced its AI into Google Docs?

The places I write the most are 1) in Google Docs, 2) in Gmail, and 3) in forums.

I use the Google Chrome browser as well.

Thus, isn’t it likely I will be able to do my work without Grammarly very soon?

What about the AI tools you can integrate into your Gmail or Outlook?

When Gmail (and Outlook) comes with generative AI native in the app, why would you even need or want a third-party tool?

One search on Matt Wolfe’s Futuretools, and you can see all the casualties…

Tools that do transcripts from online meetings face an uncertain future.

Microsoft owns the largest online meeting tools with Teams and Skype.

Google Meet is another big player.

With transcripts native in these tools, why do you need a transcribing tool like Firefly AI or MeetGeek?

Sure, there are other tools like Zoom or my friend's company Whereby. But I think it is just a matter of time before they add transcriptions into their tools as well (it is just an API away, after all).

What about the companies that do slides and presentation decks generation using AI? Will anyone pay for them when they come as a free feature inside Google Slides?

Or what about the tools that help you with your copywriting and integrate into Google Docs? Now Google will have that for free natively.

Or what about those AI tools that integrate into Google Sheets and help you generate formulas and nice graphs using AI? Well…

Like I said, a blood bath!

I think this move by Google will affect ChatGPT as well.

If I can get all my generative AI needs met inside my preferred tools, I do not need to go to/use ChatGPT (or Bing).

Sure, I understand that ChatGPT can do so much more than what happens inside Google Workspace, so honestly, I will have both.

However, for many less “advanced” users, I think their AI needs will be covered inside the Google tools.

So I cannot but wonder if this will have some sort of effect on ChatGPT.

And another thing…

If both Google and Microsoft have all these capabilities built in, will this not affect how developers think about developing new tools?

If these powerhouses of tools have all this power and capabilities, fewer tools may be needed overall, as Google Workspace and Microsoft Office will “do it all”?

In short, this AI race may kill a lot of smaller tools…

And we have not even talked about other giants like Apple, IBM, Tesla, Amazon, Meta, Salesforce, Yahoo, Adobe, Cisco, and all the Asian and European ones I don’t know about…

What do they have coming down the pike?

Wrapping up…

It is clear by now that Microsoft will make a big dent in Google’s search dominance with GPT-4 and Bing.

At the same time, it feels like Google might be able to make an equal impact on Microsoft’s dominance in enterprise/business tools with its AIs.

Wouldn’t that be ironic?

My mind is blowing my mind 🤯

Earlier this week, I was recommended a tool called my mind.

The tool promises to remember everything without you needing to organize anything (the AI does it for you).

I do a lot of research.

I now have 2 newsletters I write every week.

But I also do a lot of research for my businesses.

I use a combination of Notion, Google Docs, and Google Sheets to organize it all.

And I must admit, it is a pain in the butt.

Despite my best efforts, the sheer volume of things I save from “a million sources” creates a mess.

When my friend Stewart from Podcast Hawk introduced me to my mind, it was love at first sight!

I am too new to this tool to give you all the details or experiences, but so far, this seems to be the answer to all my prayers…

And since they use AI to organize it all, it was obvious that I had to share this find with you all.

I feel my words cannot do this tool any justice, so watch the video below and be blown away too...

(This is not a paid promotion, they don’t even know I am writing this).

The robots are coming to football (soccer)

I found this video about ARTEMIS, a football-playing humanoid robot developed at UCLA.

Although it walks as it shat itself, it is interesting to see how stable it is and how robotics are progressing.

It is now ready for the World cup for robots, the RoboCup 2023 in France.

Quick bites - interesting AI news

There is so much happening this week in AI it is hard to keep up. Here are 5 stories I found extra interesting.

1. GPT-4 is coming to LinkedIn (another Microsoft company)

2. Microsoft did not have the required infrastructure when investing in OpenAI back in 2019. They made a “leap of faith” that is paying off today.

3. OpenAI is not so “open” anymore. The GPT-4 release is the most secretive release ever, marking its full transition from a nonprofit research lab to a for-profit tech firm.

4. Duolingo launches a subscription with GPT-4 integration (Duolingo was one of OpenAI’s test partners for GPT-4)

5. Spotify introduces DJ, an AI-powered DJ, for your listening pleasure.

Shameless plug - my new newsletter

This week I launched a new newsletter called StoryBehind.

The pitch is simple: Learn the stories behind the most iconic songs in 5 min or less.

It is short and sweet and sent out on Tuesdays.

If you sign up, I will send you the story behind one of the biggest songs in history, Bohemian Rhapsody by Queen.

Sounds interesting?

You can subscribe here: https://storybehind.co/ 

Did you enjoy today’s newsletter?

How about that Spiderman, huh?

Care to share the newsletter on Twitter, Facebook, or LinkedIn?

It would be greatly appreciated!

If you have any questions about AI or any feedback, just hit reply or tweet me @thomassorheim 

AI’ll be back!

Thomas

PS! 
Why did the large language model go on a diet?

.

.

.

Because it was gaining too much "byte"!

This is not the end. It is where the fun begins!