😮Meta, OpenAI edge closer to AGI

Reading time: 5 minutes  

 Newsletter | News | Tools | Grow My Tool  

Today we will discuss-

• 📢Biden, Kishida announce joint AI research initiative

• 🙌OpenAI, Meta to release AI models with reasoning ability

• 👨‍💻MIT develops automated AI red teaming technique

• ⚙️9 amazing AI tools you might not have heard of

All this and more - Let's dive in!

👩‍🍳What’s cooking in the newsroom?

Biden and Kishida announce $50M joint AI research program funded by Amazon, Nvidia 

🍵News - U.S. President Joe Biden and Japanese Prime Minister Fumio Kishida have brought Amazon and Nvidia on board to finance a new joint artificial intelligence research program, as the two nations seek to enhance collaboration in the rapidly advancing field of technology.

According to a senior U.S. official, this $50 million project will involve joint efforts between the University of Washington in Seattle and Tsukuba University near Tokyo.  Additionally, the two countries are planning a separate joint AI research program between Carnegie Mellon University in Pittsburgh and Keio University in Tokyo.

Notably, this push for increased research in AI comes as the Biden administration evaluates a series of new regulations designed to mitigate the risks associated with the technology. 

🍜What's more? During a roundtable with business leaders in Washington on Tuesday, Kishida encouraged American executives to boost investment in emerging technologies, including artificial intelligence and semiconductors, within Japan.

He emphasized that these investments would contribute to Japan's economic growth, which would in turn lead to increased investment from Japan into the U.S., ensuring a mutually beneficial flow of investments.

OpenAI and Meta set to launch AI models with reasoning capabilities

🤓News - According to a recent report, OpenAI and Meta will soon release advanced AI models with the ability to reason.

Both Meta and OpenAI claim that their new AI models represent significant progress in giving machines abilities that surpass human cognition, especially in areas like reasoning and planning. Executives from both companies have confirmed they'll soon launch these new large language models, marking a big step towards creating Artificial General Intelligence (AGI).

🔜What's coming next - Meta has announced that it will launch the third version of its LLaMA model within the next month. 

Joelle Pineau, Meta's VP of AI research, stated that the company's focus is on enhancing these models to move beyond basic conversation abilities towards genuine reasoning, planning, and memory capabilities. Pineau emphasized that their ongoing efforts include equipping these models with the ability to not just communicate but also to think, plan, and remember things like humans do.

On the other hand, OpenAI hasn't specified the exact release date for their upcoming AI model, tentatively named GPT-5, but has indicated that it will be released in the near future.

Brad Lightcap, OpenAI’s COO, highlighted that while today's AI systems are proficient in specific tasks, they are still limited in their capabilities and only scratch the surface of what is possible. To this end, he stated that the upcoming version of GPT will excel at handling challenging tasks like reasoning, indicating a move towards AI systems capable of addressing complex tasks more effectively.

MIT researchers develop automated red teaming technique to make language AI models safer 

👨‍🔬News - MIT researchers have developed a method using machine learning to enhance the red-teaming process for large language models like ChatGPT.

Instead of relying solely on human testers to identify toxic prompts, the researchers' technique enables a red-team model to autonomously generate diverse prompts that get a broader range of undesirable responses from the chatbot. This innovation aims to address safety concerns associated with language models potentially generating harmful or unsafe content.

🤔How did they do this? They achieved this by training the red-team model to be inquisitive when crafting prompts and to prioritize unique prompts that trigger harmful responses from the target model.

🧐Did it work? Yes. Their approach surpassed human testers and other machine-learning methods by producing a greater variety of prompts that provoked more toxic responses. 

Moreover, this method not only enhances the range of inputs tested compared to other automated techniques but also reveals harmful responses from a chatbot even if it has been designed with safety measures by human specialists.

🙆🏻‍♀️What else is happening?

• New bill would force AI companies to reveal use of copyrighted art

• Intel reveals details of new AI chip to fight Nvidia dominance

• Microsoft to invest $2.9 bln to expand AI, cloud infra in Japan

• Apple AI paper details how its multi-modal ‘Ferret-UI’ LLM can interpret phone UIs

• Underwater drone equipped with AI helps researchers study the great barrier reef

💅AI life hacks

• Personalized speech writing - Public speaking can be nerve-wracking, but using AI, you can generate a personalized speech for any occasion easily. You simply need to provide the AI model with some key data and the information you want to include in the speech, and it will take care of the rest.

• City travel guides - If you’re visiting a new city for the first time, AI can help you find the best places to visit, and even provide information on the best times to see them. This can help you plan your itinerary and make the most of your time in the city.

• AI for academic research - Make use of AI-powered search engines for your academic research. AI-powered search engines like Consensus answer queries using data solely from research papers. This can save you hours of filtering through irrelevant pages and is perfect for literature reviews, or obtaining blog citations.

👩🏼‍🚒Discover mind-blowing AI tools

1. Glarity - A ChatGPT Summary extension for YouTube, Google, Twitter, and any webpage ($12.49/month)

2. WisdomAI - Allows users to upload content to create a conversational chat powered by GPT-4 ($7/month)

3. Listnr - Allows users to create realistic voice overs from text with over 1000+ voices ($19/month) 

4. Linguix - A writing assistant tool that helps improve content in multiple languages ($15/month)

5. Competely - An AI-powered competitive analysis tool that helps users quickly gather information about their competitors ($39 one-time payment)

6. Bluebuddy Caption AI - An AI-powered tool that generates engaging captions for social media posts ($0.60/97 credits)

7. Tripmix - An AI travel planner that crafts personalized trips based on individual preferences (Free) 

8. Securewoof - An AI-powered malware scanner that checks executable files for maliciousness (Free) 

9. Softr AI - An AI-powered tool that generates web applications from a single prompt (Free, $49/month)

🐦Tweet of the day