Antonym: Walking with AI Edition
Too much launch.
Dear Reader
Walking Around and Talking to AI
This week, I wrote the newsletter while walking around and talking to ChatGPT. Walking around while interacting with ChatGPT alleviated feelings of claustrophobia and sluggishness, I sometimes get from sitting at a desk for two long.
Before I went out I uploaded a previous newsletter as a style reference and the set of notes I keep through the week about things I might include. It understood the desired tone, style, and structure and offered suggestions that were largely useful.
As we walked, I discussed the newsletter’s outline and ideas aloud, effectively I had ChatGPT interview me with my notes as the structure. After I’d spoken, ChatGPT would paraphrase or interpret what I was saying, which allowed me to reflect on its response.
I’ve edited and re-written a great deal of what I co-wrote with ChatGPT walking round my local park, but I’ve kept some of the interview style sections, as I think they work well.
The week of big AI announcements
What a mess of announcements we have had this week. In the pre-social media, pre-AI age, you’d have had some comms people writing their resignation letters if you’d suggested the launch plan as either:
a. OpenAI: We’ll do an hour long presentation of a product named very similarly to the previous product but with a lower case “o” that will look like “40” when typed in a lot of fonts, and then we will make this and the current product free to everyone, but not all the time and won’t make it clear to current paying customers why why should keep paying their monthly subscriptions. We will then issue a series of clarifications, qulaifications and new announcements through the week via a combination of blog posts, tweets, messages in app and emails. None of these will have a consistent message and may contradict each other.
b. Google: We will hold a global press conference / consumer / developer announcement live for three hours launching more than 100 things at once. Some are ready, some will come out we’re not sure when and others will be available in every market except the UK or EU unless people use a VPN to access them. We will announce huge upgrades to our paid products but not make them available to millions of the paying customers but we won’t directly address that or say when we we will launch in those territories.
OpenAI also saw fit to “bury the bad news” of two of its most senior advocates for AI safety (a.k.a. alignment) leaving the company.
All that aside, there was treasure in those heaped info-dumps. Let’s take a look. I’m going to talk here and now, by which I mean things you can do today with the new features, and then some of the things that are coming next.
Is ChatGPT 4o (the o is for omni) any good?
ChatGPT 4o is a faster and incredibly capable upgrade from ChatGPT 4. It remembers more of the conversations it has had with you and directly import files to work on from Google Drive and Microsoft OneDrive (as before you can upload PDFs, CSV files and the like).
What does that mean? Complex fiddly reports, estimates and plans can come together a lot quicker. I connected my company’s financial documents and had it run a basic analysis and create some graphs in a couple of minutes. I’m used to working with different AI systems, but I was shocked how useful this was as part of my work.
Is it all free now?
ChatGPT 4 for all for free means everything that was true last week is still true, but more so. More people will be using it in the workplace whether they have permission or not. Work - from school essays to job applications and academic papers – will be harder to spot as AI-assisted.
The benefits of paying for a company licence (Teams or Enterprise) is going to be about security and being able to share tools among colleagues. Also, I think, you will only be able to build custom chatbots on premium accounts (OpenAI calls them, confusingly GPTs) and most businesses will want to tailor their own.
Is it is an even playing field now?
No. Companies that have already started AI literacy, policy, and innovation initiatives will gain the most from these updates. These organisations are well-positioned to incorporate new features and enhance their competitive edge. They will find that their bots are faster and easier to use, previously invested resources become more broadly useful, and new possibilities emerge beyond earlier limitations.
What specific benefits do you see for businesses that are already advanced in their AI integration efforts?
Organisations that are already engaged will find that their bots are faster and easier to use. Prompting class names that they’ve already invested in are now useful to more people. Tasks that weren’t possible before are now feasible, enabling organisations to push beyond previous limitations and explore new applications of AI technology. They also experience the benefit of understanding new opportunities faster due to their investment in AI literacy, resulting in a cumulative ROI over time.
How should businesses that are not as far along in their AI journey prepare to catch up and leverage these updates?
The first step is to immerse a leadership team or pilot group in AI through hands-on workshops (like our AI-B-C). This team should tackle both small tasks (e.g., meeting notes, emails) and larger strategic challenges and build their own solutions. Gradually roll out AI tools across the organisation with different access levels tailored to users’ needs. Increased AI literacy across the organisation ensures innovation at every level.
For example, at a recent client AI day, I sat a Chief Operating Officer and a finance executive in the same break-out group, both building custom chatbots in Chat GPT Teams to streamline their tasks. The ops boss made a contract screening tool to help spot the things they looked out for in legal documents, and the finance person simply wanted to save time by extracting details from PDF invoices to enter in a spreadsheet.Both reported saving significant amounts of time and frustration using their self-made tools.
That company’s staff built about 80 custom GPTs that day.
Are there any particular strategies you would recommend for ensuring security while maximising the use of these new AI capabilities?
During the initial phases, security can be managed by using secure platforms like ChatGPT Teams or Enterprise, implementing two-factor authentication, and educating staff about the risks and best practices. Regularly clearing chat histories and avoiding the input of sensitive data also help mitigate risks. Additionally, ensure bots and conversation links are shared internally and not published publicly to maintain control over access. [We give clients security guidance but defer to the final verdicts of their security experts, of course.]
What about the risks associated with generative AI, like deepfakes?
Increased AI literacy actually reduces the risk associated with generative AI. Familiarity with AI tools and understanding their potential misuse makes users more aware and capable of recognising and mitigating these risks. Regular updates on AI security developments are essential to keep the organisation informed and prepared.
For more on the AI announcements, take a look at our sister newsletter BN Edition.
Multimodal marvels
There’s more in the works. I’ve not got into the multimodal elements of Google and ChatGPT, or the realtime chat (my interview by ChatGPT above was using the old chat function) which will be coming out in the next coupe of weeks.
Also, take a look at these two incredible videos from OpenAI which show how these tools and our use of them will be evolving in the coming weeks.
One is the founder of Khan Academy and his son with ChatGPT talking them through a maths problem, and the other is the two GPTs talking to each other. And then singing.
Seriously, if you’ve not seen them yet, have a watch.
That’s all for this week!
I need to lie down.
Thank you for reading, as ever, if you liked it, share it with someone. Thank you.
Antony