ChatGPT's latest feature drop, plugins of the week, developer info now available

🔥 Plugins of the Week

There are now 849 ChatGPT plugins in the plugin store. Let’s dig right into a few of the 36 new ChatGPT plugins released in the past week.

‍

‍

1. ‍Paperpile - Supercharges research by searching millions of scientific papers in seconds.

Prompt examples:

"Search for papers discussing the use of AI in healthcare."

"Find abstracts related to the impact of global warming on marine life."

‍

2. ‍Convert Assistant - Offers guidance for convert.com's AB testing tool.

Prompt examples:

"Tell me about A/B testing on Convert.com"

"How can I use Convert for my experimentation?"

‍

3. ‍Solana Labs - Explores Solana blockchain data.

Prompt examples:

"Show me NFT collections matching the keyword "Monkey"

“Show me listed NFTs for the Monkey Baby Business collection”

‍

Here are the categories with most new launches:

📊 Data & Research - 96 plugins (6 new)

📣 Marketing & SEO - 40 plugins (5 new)

🔮 Miscellaneous - 94 plugins (4 new)

‍

✨ ChatGPT’s Feature Drop

OpenAI just rolled out new features for ChatGPT:

• Prompt examples at the beginning of a chat.
• Suggested replies to continue the conversation.
• Multi-file upload to analyse and get insights across multiple files (using Code Interpreter).

In the coming week, more features will roll out, including keyboard shortcuts (type ⌘ [Ctrl] + / to see all shortcuts) and minor updates like GPT-4 set as the default model for Plus users, as well as users no longer being logged out after 2 weeks.

Why I’m excited about it: This update could increase adoption from new users, which were previously confused about how to start and how to continue a conversation with ChatGPT. The multi-file upload is super handy as well – I just tested it for comparing 2 datasets and finding differences, and it worked like a charm.

‍

👨🏻‍💻 Developer Info Now Available for All Plugins

Both on the plugin store and on whatplugin, developer info including terms of service and/or privacy policy, as well as a contact e-mail is now available for every plugin.

Now, we can more easily communicate directly with developers to give feedback, ask questions, etc. on their plugins 🙌🏽

‍

💎 Why Plugins Matter (even if they sometimes suck)

Yes, I said it. While there’s absolute gems hiding in there, the plugin store has several gimmicky and plain unhelpful plugins, which is why I recently introduced a reporting feature on whatplugin.

Here’s why ChatGPT plugins are generally awesome, though:

• The top plugins can make you way more productive in ChatGPT, enabling you to build efficient workflows using real-time data.
• While sometimes slow and buggy, new plugins are interesting because they show you how companies can potentially benefit from AI assistants like ChatGPT.
• It’s essentially an experiment in developing multimodal AI systems – by using and following the development of plugins, you’re staying at the forefront of how this interconnectedness can and will work in terms of things like data sharing (be careful though) and system-to-system communication.

‍

📈 AI for Business is Booming

• AI was a hot topic in Q2 earning calls. Companies were talking less about recession, and more about AI in earnings calls in Q2, with a 323% increase in mentions of AI compared to last year, according to Bloomberg.
• Intel to integrate AI into every product. On their Q2 earnings call, Intel’s CEO told investors about their plans to “build AI into every product that we build”. First up is the company’s Meteor Lake processor dropping later this year, featuring a built-in neural processor for machine learning. Intel’s recent move means we are likely to have more AI operating locally (doesn’t rely on cloud services) on our computers soon.
• Common Issues and opportunities for AI in business. Andrew Ng of Deeplearning.ai just released a short, super-interesting read on the common AI-related issues that businesses are currently facing, and a practical approach to exploring opportunities in AI for any company. Some takeaways:
• The most common implementations of AI in business are currently customer service chatbots and answering questions based on documents.
• Businesses frequently ask how AI will affect their company and how to use it without giving away their data to platform vendors, while governments wonder how it will affect the labour market.
• It’s beneficial to consider not jobs as a whole, but component tasks of those jobs, and ask which of these are suited for automation or assistance by AI.

‍

🤖 Other Chatbot & AI News

• GPT-5 trademark applications hints at new features. However, Sam Altman states OpenAI is not close to a release anytime soon.
• OpenAI discontinues its AI-written text detector due to low accuracy. The tool was designed to determine the likelihood of a text being generated by AI, something which in the past has proven difficult for the many AI-text detection tools out there.
• While no AI detection tool is perfect, I like the approach of Originality.ai on transparently developing an open approach to comparing AI content detection accuracy across different detectors.
• Google Bard struggling with visual understanding. It’s now possible to prompt Google Bard with images. However, a recent study testing the chatbot’s capabilities on visual understanding found substantial room for improvement, highlighting the the gap in vision-based understanding that should be addressed in future AI models.
• Amazon is making it easier to build applications with generative AI. AWS just introduced new foundational models to its Bedrock platform. Models available at this point include AI21 Labs, Anthropic, Cohere, Stability AI, and Amazon – accessible via a simple API interface to quickly and securely build AI applications. There’s an ongoing AI race and partnership forming among the biggest players in cloud computing; Microsoft Azure currently offers access to OpenAI’s models GPT-4 and DALL·E, while the Google Cloud platform gives access to its PaLM large language model and Imagen image generator.