99 | ChatGPT 4o is 4u ... for now!
In the theme of Rose, Bud, Thorn, OpenAI’s announcements surrounding the new GPT 4o model are beautiful buds. The question is whether they’ll blossom or wither in the shade of the tech giants, Google and Microsoft.
Let’s start with what is available for use today. My take?
ChatGPT 4o is for you
The “o” in “4o stands for omni, highlighting the work OpenAI did to combine their different models (general LLM, image generation, etc.) into a single model. This is a huge engineering feat, catching up with Google’s Gemini. What does it mean for users like us? It should provide faster responses to prompts, short-circuiting the layer in ChatGPT that would interpret a prompt and decide which model(s) to submit it to.
Beyond that leap forward in fundamental model architecture, the new ChatGPT 4o includes many new features, some available now and some to be rolled out in the coming days/weeks. Check out the OpenAI announcement here, which includes some compelling video demos.
For ChatGPT Plus Subscribers
If you’re paying $20 per month for ChatGPT Plus, you just got a nice boost. Just be sure to choose ChatGPT 4o at the top of your screen. You still can use ChatGPT 4 or 3.5, but 4o is the way to go…
The 4o model is simply much, much better. This has been observed subjectively by many AI thought leaders and podcasters. But also, it’s fairly likely that it’s been validated through what’s speculated to be the early leak of the model on the LMSYS Chatbot Arena (mentioned in this recent StrefaTECH article), where the “im-a-good-gpt2-chatbot” mystery model soared to the lead. That the model is GPT 4o has been “confirmed” by Ars Technica. Regardless, it’s likely that the responses to your queries will be much better than even the state-of-the-art ChatGPT 4 you’ve been using.
Internet browsing is way better. OpenAI is clearly working toward trouncing Perplexity and giving Google a run for their money in search. ChatGPT has used “Browse with Bing” (Bing is Microsoft’s web browser), but with ChatGPT 4o, the results for questions about current/recent events are much better and include direct citations, usually from trusted news sites.
For example, ask:
What's happening in the Stanley Cup playoffs?
What's the latest in the Gaza conflict?
DALL-E can do much more. If you want consistency of characters or other image elements across multiple prompts, DALL-E finally is able to meet your need. For characters, a key tip is to name the character.
For example, try these two prompts:
Image of a young woman named Sarah with long, wavy brown hair, blue eyes, wearing a green dress and a silver necklace, standing in a sunlit meadow with flowers
Sarah, the young woman with long, wavy brown hair and blue eyes, now wearing a blue winter coat and scarf, walking through a snowy forest
Charting and analysis is vastly improved. The original ChatGPT “code interpreter” model which was renamed “advanced data analysis” for awhile, is now simply a part of the new GPT 4o model, but has taken huge strides forward in both data analysis and charting. Try this:
Upload a spreadsheet (deidentify any sensitive data first)
Ask ChatGPT 4o questions like:
Analyze this spreadsheet for interesting insights
Perform a formal statistical analysis of the relevant data
Generate some charts and visualizations that illustrate these findings
Compile these into a PDF
There’s more available now and much more coming, particularly in the areas of interactive conversations and image interpretation. See the OpenAI announcement or just search on YouTube for the demos.
Not paying for ChatGPT? Watch for 4o to come to you soon
If you’re not paying for ChatGPT, first, I hope you’re using some other chatbot (e.g., Claude, Copilot, Gemini), because they’re all far superior to ChatGPT 3.5, which is the current free offering from OpenAI. This is about to change, though.
OpenAI’s announcement include a huge change—soon, their free model will be ChatGPT 4o, although its use will be limited to 16 prompts in a 3-hour period, which for many is likely to be too few.
Also, free ChatGPT 4o users will be able to use custom GPTs. This could be very valuable to organizations that want to use ChatGPT to create bespoke chatbots for employees and other stakeholders to “chat with” their complex documents, such as HR policies.
But this may just be “for now”
While OpenAI continues to offer the most feature-rich app in ChatGPT and the leading model in GPT 4o, generative AI has much more to offer than going to a general-purpose chatbot to have conversations. In particular, as Google and Microsoft integrate AI into their products, which are used by millions daily for both business and personal use, it’s likely that a day will come when AI is just part of what we do in our go-to apps. This could render these separate apps like ChatGPT redundant.
If (or when?) that day arrives, it’s a win for users. Being able to tap into the power of generative AI from within the set of apps we’re already comfortable with, pulling in data that we already store in various places (Google Drive, Notion, CRM systems, etc.) is the next nirvana.
While this doesn’t spell the likely demise of OpenAI—they’ll continue to provide the underlying AI engine to many software vendors—it may be that the glory days of ChatGPT are limited.
But for now, it really is an exciting time! The leaps forward are hard to catch up with, but once we do, they’re amazingly helpful (and sometimes just plain fun!).
As you ponder how to invest your time, I continue to exhort you to play with AI, use it any time you have a question, and hone your intuition about when it fits and when it doesn’t. (Hallucinations are still an Achilles heal.) And as you plan your time and decide when to trust what “the AI” tells you, be sure to…
Make Good Choices