- OpenTools' Newsletter
- Posts
- 🚨 GPT-5 coming soon?
🚨 GPT-5 coming soon?
PLUS: Central banks embrace generative AI
Reading time: 5 minutes
Can you help us, please?Last week, we asked if you'd be interested in a free prompt guide, and the response was incredible! We're thrilled to share that we're already working on it. Now, we need your help to make it as useful as possible. |
Today we will discuss-
🏋️GPT-5 in training !!?
🌐Central banks embrace generative AI
🤗GPT-4 shows compassion
⚙️9 amazing AI tools you might not have heard of
All this and more - Let's dive in!
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/9b3f3bf6-173d-475e-959a-f1146ad29e2d/animated-worm-image-0130.gif)
👩🍳What’s cooking in the newsroom?
GPT-5 in training?
The OpenAI Board announced today that they are putting together a “Safety and Security Committee”. The committee will be led by Bret Taylor (Chair of the Board), Adam D’Angelo, Nicole Seligman, and none other than the lizard king himself Sam Altman (CEO).
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/ff0c1dd2-d0a2-4abd-bc0e-2579cbd20bad/robert-california-lizard-king.gif?t=1716914736)
They’re tasked with making security recommendations on all OpenAI projects within 90 days.
But that’s the boring part.
The interesting part is that OpenAI revealed in this announcement that they are officially training their new frontier model (which we can only assume is GPT-5, right?).
My opinion - I think this is all corporate theatre in advance of a major announcement (hopefully GPT-5). So, my team will build with the assumption that there will be a massive update released in the near future.
You can read their announcement here
But what about GPT-4o?
GPT-4 was launched over a year ago in March 2023 - and earlier this month they released GPT-4o - a faster version of GPT-4 with vision and voice built in.
From my experience (and that of my twitter friends) - in terms of performance, GPT-4o is much worse when compared directly to GPT-4 on how well it completes a task. However, GPT-4o is much quicker, and is about 50% cheaper.
So even though it underperforms GPT-4, its a much needed upgrade for basic tasks that I used to give to GPT-3.5.
That being said, I think they are aware that they’ve milked GPT-4 dry, and are finally gearing up for the GPT-5 release.
Generative AI gains traction among central banks for enhanced cybersecurity: BIS report
🏦News - The Bank for International Settlements (BIS), which includes 63 central banks and monetary authorities, sees significant potential in the widespread use of generative artificial intelligence.
In a survey of 32 central banks, the BIS found that 71% are already using generative AI, and 26% plan to implement it within the next one to two years for cybersecurity purposes.
The BIS anticipates that all its members will eventually adopt generative AI to improve internal cybersecurity. This aligns with findings that central banks already using generative AI have found it more effective in detecting cyber threats than traditional methods.
👨🏻💻What's more?
Generative AI tools have sped up banks' response times to cyberattacks and improved their ability to detect suspicious trends and anomalies.
Central banks unanimously believe that generative AI tools can eventually take over routine cybersecurity tasks from staff. The BIS expects this shift to "free up resources" for other initiatives.
🤔Are there any concerns though?
The expenses associated with adopting generative AI tools are the main concern for central banks.
Furthermore, the BIS report highlighted that there are also concerns related to social engineering, zero-day attacks, and unauthorized data disclosure.
OpenAI's GPT-4 excels at empathy test that most human doctors fail
😇News - A recent study published in a scientific journal found that OpenAI's latest language model, GPT-4, has made a significant breakthrough in mimicking human communication by showcasing a level of empathy.
The study tested two chatbot models, GPT-4 and GPT-3.5-turbo (ChatGPT) to see how well they could perform on the U.S. Medical Licensing Examination, which assesses qualities like empathy, communication, professionalism, and ethical judgment in future doctors.
🤓What were the key findings?
Among the two models, GPT-4 excelled on the test with a 90% correct answer rate, outperforming ChatGPT, which achieved a 62.5% accuracy rate.
GPT-4 displayed higher confidence by not revising its responses, whereas ChatGPT altered its initial answers 82.5% of the time.
Moreover, when ChatGPT corrected its original wrong responses, it successfully provided the correct answer in 53.8% of those cases.
🤷🏻♀️Why is this important? Chatbots and AI tools are already being used in hospital medical services to reduce burnout among healthcare professionals. Their potential to replicate human empathy opens up new opportunities for them to enhance softer clinical skills and be used in psychology, therapy, and telemedicine.
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/6b228770-13e3-452b-993c-10d93372679b/animated-television-image-0048.gif)
🙆🏻♀️What else is happening?
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/deb5fcaa-6788-452c-8514-169064927fa2/unnamed.gif)
💁🏻♂️Prompt engineering tips
Prompt Pruning: Remove unnecessary or redundant information from your prompts to help the model focus on the most relevant aspects of the task.
Prompt Gradation: Gradually increase the complexity of your prompts to gradually challenge the model and improve its performance over time.
Repetition: Repeat key information or instructions in your prompts to reinforce the model's understanding and improve consistency.
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/61b8a603-ad02-4725-b0d1-ad081b898a02/animated-tool-image-0190.gif)
👩🏼🚒Discover mind-blowing AI tools
Bugasura - A bug tracking tool that enables users to report, track, and close bugs faster (Free, $5/month)
Comicai - An AI-powered tool that allows users to create comics easily ($15/month)
Cookii - A cooking assistant that allows users to search for recipes based on their dietary restrictions and cuisine preferences (Free)
ZeroBot - A voice-enabled chatbot that allows users to have conversations with a computer friend that talks like a real person (Free)
Tripmix - An AI travel planner that crafts personalized trips based on individual preferences (Free)
Magnific - An AI-powered image upscaler and enhancer that increases the resolution and detail of images ($39/month)
MurfAI - An online voice generator tool for your videos, presentations, and brand commercials ($26/month)
Competely - An AI-powered competitive analysis tool that helps users quickly gather information about their competitors ($39 one-time payment)
Mockey - Provides users with over 1000 product templates to create product mockups easily (Free)
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/f21db1b1-c413-45d6-8168-d70c01f7815e/animated-cycling-image-0058.gif)
🥴Unusual AI use cases
Funny Cat Videos: AI can create hilarious and entertaining cat videos by analyzing thousands of existing videos and generating new ones with unexpected and amusing twists.
Phone Ringtones: Use AI to create unique and unusual phone ringtones, like a symphony of animal sounds or a mashup of classic movie quotes.
Fortune Cookie Messages: Use AI to generate fortune cookie messages with amusing and playful predictions, guaranteed to spark laughter and bring joy to diners as they enjoy their meal.
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/18dda003-9a8a-4f8a-b690-f92c4d68b48d/animated-thank-you-image-0091.gif)
Can you help us, please?Be our mentor and let us know your take on today's newsletter. Share your honest and detailed thoughts or just drop a Hi. YES! We personally read all the messages. |