Howdy, wizards. As April comes to an end, I’m rounding up the latest developments in custom GPTs as well as the best new ones for you to try. Check the GPT of the month for an easy use case of GPTs you can apply immediately.
Also, ICYMI – here’s my article on the best tools for GPT builders.
What’s brewing in AI / GPTs roundup April 2024
- Moderna’s success with custom GPTs
- The GPT store is lacking some moderation
- GPTs – April roundup: Top newcomers and GPT of the month
Dario’s Picks - GPTs special
1. Moderna’s success with custom GPTs
.jpeg "GPTs roundup: April 2024 | whatplugin blog (1)")
Moderna is using ChatGPT Enterprise across their organization, in particular the ability to create custom GPTs. It’s making them more efficient and accurate in their work, while maintaining privacy.
In less than 2 months, employees have created 750 custom GPTs. Building custom GPTs with ChatGPT Enterprise is helping Moderna become more efficient and acc
urate in their work, and helping employees focus their attention on higher-level tasks, without compromising privacy.
Here’s some popular GPTs the company is using internally:
- Dose ID: uses the Advanced Data Analysis feature to review clinical data and visualizes large datasets, aiding the team’s decision-making.
- Contract Companion GPT (pictured above), creates a readable summary of any contract.
- Policy Bot GPT: gives employees quick answers about internal policies instead of needing to search hundreds of documents.
- The brand team has a GPT to help create slides for quarterly earnings calls and one to translate biotech jargon into approachable language for investors.
Context
Moderna is a pharma and biotech company most well-known for their covid vaccine. They’ve partnered with OpenAI for over a year, and ChatGPT Enterprise is evolving how the company is using AI internally. They’ve put down a team of experts tasked with driving AI adoption (using AI-assisted trainings, internal AI forums, prompt contests, and more innovative ways to achieve this).
Initially, they had developed an internal chatbot, mChat, which was widely used in the company; however, with the launch of ChatGPT Enterprise, the company decided to just switch to this platform since user testing indicated a better experience with ChatGPT’s interface.
Why it matters
Very promising to hear about these internal use-cases for GPTs. I think the company played it well by switching from a custom chatbot directly to ChatGPT – a lot of the times UX is more important than the performance aspect.
Internal GPT usage is the arena where I think the best custom GPTs is being made these days, not in the public GPT store. A problem with many public GPTs is that they are simply thinly veiled shells for premium services. On the other hand, employees making GPTs for internal use are incentivised by making their team’s work easier, they have strong domain knowledge and real problems to solve, and are not doing it for external marketing reasons.
I’m hopeful that the public GPT arena can become a highly useful app store for people and companies to readily adapt. For that to happen, stronger moderation is needed from OpenAI (more on this below).
2. The GPT store is lacking some moderation
OpenAI’s GPT store continue getting negative press regarding their lack of moderation. Tech Crunch did an article on it last month, and picked up in Andrew Ng’s newsletter last week.
There’s lots of GPTs currently in the GPT store that are most likely in breach with OpenAI’s own terms and guidelines:
- GPTs purported to jailbreak ChatGPT: If they do what they say they’d go against OpenAI’s safeguards.
- Text “humanizer” GPTs: These could possibly be in breach with OpenAI’s ban on GPTs that encourage academic dishonesty.
- GPTs including trademarked names or characters. There are lots of these, and supposedly the store prohibits using third-party content without authorization.
- GPTs impersonating real-life people e.g. Elon Musk, Jordan Peterson, etc.
The GPT Store’s low barrier to entry is a boon to well-meaning developers, but it may encourage less responsible actors to take advantage of lax moderation. We applaud OpenAI’s willingness to execute an ambitious vision and hope it finds a workable balance. - Andrew Ng, The Batch
Why it matters
OpenAI is having a similar problem to Google search these days – easy-to-produce spam undermining the user experience.
Manually moderating the thousands of GPTs created daily may not be feasible at this point. But I think having a small team actively curate the ≈100 GPTs featured in the store (and check all GPTs indexed in the search) is a low-hanging fruit for boosting the credibility of the store.
GPTs - April roundup
whatplugin’s top 100 - April newcomers
GPTs that entered whatplugin’s top 100 rank in March. How rankings work (also below).
How I’m ranking these GPTs
%20(1).png "GPTs roundup: April 2024 | whatplugin blog (2)")
In case you were wondering – rankings are based on an algorithm that takes several aspects of GPTs into account. I'm trying to be as objective as possible, although manual reviews can also affect the ranks.
I’ve provided a bit more detail in the image below:
My goal is to make a system that surfaces useful and safe GPTs, while making it difficult to manipulate the ranks – but please keep in mind usage is at your own risk.
GPT of the Month
💬 10K+ ⭐️4.2 out of 300+ ratings
For legal research, contracts, law resources, law terminology, law & legal education.
.png "GPTs roundup: April 2024 | whatplugin blog (3)")
Dario: Using ChatGPT to make sense out of legal docs is a life hack (although, not legal advice…)! There’s several GPTs out there like this – a pre-prompted version of ChatGPT designed to simplify legal matters. Try uploading your insurance policy, employment agreement, etc. Ask it to summarize it, ask questions about it, etc.
'); $('.blog-post-rich-text').html($htmlContent.html()); $('.blog-post-rich-text a').attr('target', '_blank');});
