[Just the Signals #9] ChatGPT Gets Its Biggest Upgrade Since May 2024

Happy Friday!

This is Just the Signals:
The edition that gives you the most important AI updates from this week.

There is so much AI news these days.
And TBH, most of it is noise and hype.

It's not framed for people in non-technical or corporate roles.
It's focused on small updates, instead of big picture trends.

It's no wonder people are confused, overwhelmed, or believe it is overblown.

So, I've combined my corporate lens with my interest in AI.
I ruthlessly follow the latest news and reports, so you don't have to.

My goal is to give you one AI source you can trust.
One that puts you ahead and in the know.
One that also saves you time and effort.

You're busy, and I'm here to help.

So, I urge you to spend just a few minutes reading this each week.
I promise you'll have an advantage over people who don't take action.

Without further delay, here's what you need to know this week:

1 / ChatGPT Gets A New Model (GPT-5)

OpenAI released its new ChatGPT model (GPT-5) on Thursday.
It's already started rolling out to Free, Pro, Plus and Team subscribers.
Enterprise and Education will get it next week.

In case you're not sure what this means, this is a BIG deal.

ChatGPT is the most used AI platform by far (over 75% traffic market share).
Majority of those users have been using their default GPT-4o model.
This model was released May 2024 (with upgrades since).

Newer models (like o3 - reasoning and GPT-4.5 - writing/EQ) have been available.
But you need to select these manually to use them.

It's been more than a year since their latest release.
AI has come a looonnng way since.

Now GPT-5 will become the default model for ALL users.
Say goodbye to GPT-4o and other legacy models.

The masses get an upgrade and will see how far AI capabilities have come.

Here are a few things to know about this release:

1. GPT-5 is the top-ranking model in the world

• For now (also notice GPT-4o isn't even on the list above)

2. You'll be able to have 3x longer convos without AI going haywire

• The context window increased from 128K tokens to 400K tokens

3. It's a reasoning model & much more capable than GPT-4o

• Better critical thinking, problem-solving, and decision-making
• Watch as it "thinks" through how to best respond to your prompts

4. They've massively improved hallucination issues

• OpenAI claims it's 45% less likely to hallucinate vs GPT-4o (80% vs o3)

5. It is ready to be used with tools

• GPT-5 will integrate with Google apps (Gmail, Calendar, etc.)
• It's designed to be used as an agent and connect with other tools via API

6. It's cost-efficient!

• For those building custom apps, API rates are cheaper vs other top models

7. We're getting a few other goodies

• The ability to change personalities (professional, supportive, sarcastic, etc.)
• Paid users can change the color of chat bubbles
• Improved memory functionalities for a more personalized experience
  

Now it's time for us to test it out!
Time and experience will tell just how GPT-5 stacks up vs competitors.

Minimally, expect the masses to get a surprising upgrade.

2/ It Was a Big Week for Other Model Releases

There had been whispers that OpenAI would be dropping GPT-5 early August.
So, what did competitors do?
They had smaller releases of their own!

1. Anthropic released Claude Opus 4.1

• An upgrade to the world's top coding model
• As I wrote about earlier this week, Claude holds the number 1 spot among enterprise AI users - they were looking to hold onto their edge!

2. Google released Genie 3

• A new model that creates interactive, playable 3d worlds (think gaming)
• Environments are created via text and have real-world physics

3. (And surprise) OpenAI also had another big drop

• They released two open-weight models (GPT-oss-120b & GPT-oss-20b)
• Their first open models released since GPT-2 in 2019
• These new models allow people to download and host AI from their servers (this has cost efficiency, privacy, and fine-tuning benefits!)

You may not be using Claude or Gemini, but it's good to keep perspective on what's going on with the other major players.

ChatGPT has the leaderboard today, but tomorrow it could flop to one of these two.

Polymarket is betting that Google has the top model at the end of the year.

3/ AI Super Users Still Live in a Bubble

A study by The Associated Press shows:

• Only 60% of US adults have used AI to search for information
• Only 37% of US adults have used AI for work tasks

The catch?
Users under 30 are significantly more likely to be AI users.

To me, the study (and above chart) show the opportunities still in play.

Last week, I wrote how jobs that have at least one AI skill requirement listed in postings average $18K higher earning potential than those that don't.

Jobs with two AI skills listed? $28K higher earning potential.

Most people are still using AI in very basic ways, if at all.
Learning skills beyond this puts you ahead.

But AI has been adopted quicker than any technology before.
ChatGPT is now at 700M users per week globally.

Now's the time to capture the advantage.

4/ A Few Legal Updates to Put on Your Radar

Everyone's favorite topic!
I'll keep it quick... don't close out here.

1. Perplexity was caught browsing sites it shouldn't have been!

• The AI search engine was called out by Cloudflare (who routes 20% of all web traffic) for scraping websites it shouldn't have been
• Allegedly, Perplexity bypassed Internet rules (robots.txt) which is a big no
• Bypassing Internet rules saying bots shouldn't scrape certain sites leads to content creators losing traffic and could set an expectation that AI companies don't need to respect owner wishes

Something to pay attention to as sites already are struggling for traffic.

2. OpenAI is still fighting The New York Times in Court

• The two are in a court case NYT alleges OpenAI illegally scraped its content to train ChatGPT
• NYT is requesting ChatGPT provide 120 million chat conversations
• OpenAI offered only 20 million chats due to privacy concerns

The case is a warning for what could happen to your chat data.
If NYT wins, this could spiral into similar data demands from other publishers.
It'll also set a precedent for the future of AI copyright and training laws.

3. Some ChatGPT Chats Were Found on Google

• People found private-seeming ChatGPT conversations in search engine results
• Turns out, people who created shareable links had the option to opt in and make their conversations public
• ChatGPT got rid of this function after it drove a strong social media reaction

Sharing this as a reminder that AI is still a very new technology.
AI companies have kinks to work through.
As consumers, we also need to be careful about what we share and select!

That's a wrap for the stuff you need to know this week. Have a great weekend!