- Mindstream
- Posts
- Medium Risk
PRESENTED BY |  |
The world’s oldest cheese, found in an Egyptian tomb, is over 3,200 years old and was likely made from a mixture of cow and goat milk. Would you try it?
Love or hate this intro? Hit reply and tell us why!
What’s in store:
GPT-4o is safe, but also persuasive.
Why AI’s future isn’t about “Digital Gods”.
Around the web: Biden reveals a plan to protect clean energy systems from cyberattacks, focusing on batteries, EVs, and smart grids.
Guidde lets you make documentation 11x faster!
Read Time: 5 minutes
OPENAI
OpenAI has released the GPT-4o System Card, a document outlining the safety measures and risk assessments carried out before the model's launch.
Unveiled in May, GPT-4o underwent rigorous testing by external security experts, known as red teamers, who identified potential risks such as unauthorised voice cloning, inappropriate content creation, and copyright issues.
The results of these tests, now public, have classified GPT-4o as a "medium" risk, primarily due to its persuasive capabilities.
The rundown:
Although most areas were low risk, the model's persuasive writing was flagged as a concern.
Internal and external evaluations were conducted, including by groups like METR and Apollo Research.
OpenAI has previously released system cards for GPT-4, GPT-4 with vision, and DALL-E 3, following similar safety testing.
Safety first, persuasion second?
The release of this system card comes at a pivotal time for OpenAI, as the company faces ongoing scrutiny over its safety standards.
This includes a recent open letter from U.S. politicians questioning OpenAI’s approach to transparency and safety.
The timing is particularly crucial given the upcoming U.S. presidential election, where AI misuse could have significant consequences.
There have been increasing calls for OpenAI to be more transparent about its safety testing and the training data used for models like GPT-4o.
In California, where OpenAI is based, legislation is being proposed that would require large AI models to undergo state-mandated risk assessments before being made available to the public.
Despite the involvement of external testers, much of the responsibility for safety evaluation still rests with OpenAI itself.
How long until our AI is convincing us to do things?
Do you think GPT-4o's persuasive abilities are a feature or a flaw? |
Vote for live results and see results + opinions from yesterday at the bottom of the email.
PRODUCTIVITY HACKS
Ever have that colleague who asks the same question repeatedly, no matter how many times you explain?
It's time to let AI handle the explanations. Meet Guidde, the AI-powered tool that transforms complex tasks into easy-to-follow, AI-generated documentation in seconds:
Transform tedious documentation into engaging visual guides.
Create video documentation 11x faster, saving you valuable time.
Document workflows, share insights, train and onboard new hires effortlessly.
Discover our new features: video intros and outros, parallel capture, additional voices and avatars plus an enhanced editor!
With our browser extension, start a capture and let Guidde generate step-by-step video guides complete with visuals, voiceover, and much more.
Join over 60,000 users who trust Guidde.
Rated 5/5 stars on the Google Chrome store.
On This Day
August 10th 1793 - The French revolutionary government opens the Louvre. It would become the world's most-visited museum.
AI Image or Real?
Our cravings today! But is this real or not?
Vote below!
The poll results and correct answer will be revealed in tomorrow’s Mindstream!
AI Image or Real? |
Mystery Link…
AI MODELS
AI companies are attracting huge investments and achieving sky-high valuations early in their development, leading some to label the industry as a bubble.
However, Nick Frosst, co-founder of Cohere, argues that while there’s a lot of excitement, calling it a bubble ignores the real value these companies are delivering.
Frosst points out that Cohere’s AI models enable new features and automate tasks that were previously impossible, demonstrating the tangible benefits of AI.
In summary:
AI firms are adding real value, despite concerns of an industry bubble.
Artificial general intelligence (AGI) isn’t on the horizon just yet.
Cohere’s strategy centres on large, adaptable language models for enterprise clients.
More than just buzzwords
Although the industry has its sceptics, Frosst is cautious about the future, particularly concerning artificial general intelligence (AGI) or human-level intelligence development.
He suggests that while AI is undoubtedly powerful and transformative, the notion of creating "digital gods" is far-fetched and unlikely to materialise anytime soon.
Cohere, founded on the research of co-founder Aidan Gomez, focuses on realistic expectations of what AI can achieve.
Their business model prioritises large, versatile language models over those that are narrowly focused on specific tasks.
Digital gods? Sounds like a straight-to-tv movie.
Mindstream Picks
Eutelsat plans to sell 80% of its ground segment infrastructure, valued at around €790 million, to a private equity fund managed by EQT Partners, the company announced on Aug. 9.
Sony President Hiroki Totoki clarified that Bungie’s restructuring aims to optimize costs and focus resources on Destiny 2 and Marathon, with other titles moving to PlayStation studios.
Ripple is developing a fiat-pegged stablecoin called Ripple USD (RLUSD), with private beta testing underway on XRP's ledger and Ethereum, as it targets the $160 billion stablecoin market.
Wall Street rebounded Friday, with the S&P 500 and Nasdaq gaining 0.5% each, recovering losses from earlier in the week driven by economic fears and overseas market shifts.
Don’t Miss - The Biden administration has unveiled a plan to bolster cybersecurity for the nation's clean energy infrastructure, focusing on protecting critical technologies like batteries, EVs, and smart grids from cyber threats. The initiative aims to ensure that new energy systems are secure by design and involves collaboration between government, businesses, and hackers.
AI Art
Our Image of the Day
Artwork submitted by Mindstream reader Monie G: “A cappuccino cup with foam spilling out”
Daily Image Prompt
A futuristic looking Tokyo
Yesterday’s Poll
“The AI Pin...”
Terrible idea, it was never going to work. - 51% ✅
Good idea, poor execution. - 49%
Your Views:
“Don't ignore client/customer feedback - and don't launch a faulty product. This is a nightmare.” - tgeerts
“Wearable tech is always fun and helpful, but you cannot mess it up.” - travis.sahr
Submit your opinions in our daily poll to be featured!
 | Have a great Saturday, Y'all. ❤️ We need your feedback to make our newsletter better. 📣 Refer our newsletter to your friends and make us feel all warm and fuzzy. 🚀 Advertise in our newsletter to reach 150,000+ founders, engineers, and content creators. |
Hey You! Would You Recommend Us? |