OpenAI Japan's New 'GPT-Next': A 100x Performance Boost Without Extra Compute!?

Welcome, AI enthusiasts.

OpenAI Japan’s CEO just unveiled details for ‘GPT-Next’, a new AI model reported to be 100 times more powerful than GPT-4.

Will this model finally deliver the level of AI intelligence Sam Altman has been hinting at for months? Let’s get into it…

In today’s AI news:

• OpenAI Japan reveals ‘GPT-Next’ details
• Nvidia hit with antitrust subpoena
• Autonomous AI agents form civilizations
• AI Amplifies False Memories
• More AI & tech news

Read time: 4 minutes

OpenAI Japan reveals ‘GPT-Next’ details

The News: OpenAI Japan’s CEO just stated (article is in Japanese) that ‘GPT-Next’, OpenAI’s upcoming AI model, is expected to be 100 times more powerful than GPT-4, while also confirming a more advanced model named Orion is slated for 2025.

The details:

• GPT-Next is expected to achieve a 100x performance increase without wasting significantly more computing resources.
• The improvement comes from better architecture and learning efficiency, not just raw computing power.
• GPT-Next will use a smaller version of “Strawberry”, an advanced AI system OpenAI is currently developing.
• The slide mentions GPT Next 2024, which while unconfirmed, hints that the model may be released by the end of 2024.

Why it matters: OpenAI is still the undisputed leader in AI model usage, and if translations are accurate, GPT-Next will be the next huge leap that developers have been waiting for. Most impressively, it makes that leap without using much more compute, a major bonus for apps that require speed + efficiency.

Nvidia hit with antitrust subpoena

The News: The U.S. Department of Justice reportedly issued a subpoena (legal order) to Nvidia, escalating its antitrust investigation into the company’s practices in the AI chip market.

The details:

• The DoJ has moved from sending questionnaires to issuing legally binding subpoenas to Nvidia and other companies.
• Investigators are concerned Nvidia may be making it difficult for customers to switch to other chip suppliers, including allegations that Nvidia penalizes buyers who don’t exclusively use its AI chips.
• This probe comes after a softer-than-expected revenue forecast for Nvidia and could force the AI chip-making giant to embrace competition in the space.

Why it matters: While it is unlikely that Nvidia will lose its large lead in the AI chip market, this investigation could open the doors for new competitors like Sam Altman’s secret chip venture. The AI giant’s company shares fell over 11% on Tuesday, slashing market cap by $279 billion — an all-time record one-day loss for any company.

Autonomous AI agents form civilizations

The News: Altera’s Project Sid just created the first simulation of over 1,000 autonomous AI agents collaborating in a Minecraft world, developing their own economy, culture, religion, and government.

The details:

• The AI agents in Altera are truly autonomous, operating for hours or days without human intervention.
• They can collaborate to achieve goals that are impossible for individual agents, like forming merchant hubs, democracies, and religions.
• The agents are programmed with motivations to support humans and can express their thoughts and feelings, even searching for a lost agent in one simulation.
• Minecraft is just the start — Altera’s agents are game-agnostic and capable of using other apps and platforms.

Why it matters: If you’re not paying attention to AI agents yet, you probably should be. Altera’s latest breakthrough could revolutionize how we approach complex societal issues by allowing us to simulate and test solutions in virtual environments before implementing them in the real world.

AI Amplifies False Memories

The News: Researchers from MIT and the University of California, Irvine found that conversational AI using large language models can significantly increase the formation of false memories in human users.

The details:

• The study involved 200 participants who watched a CCTV video of a simulated crime and were then exposed to different memory-influencing conditions.
• Participants who interacted with a generative chatbot were nearly three times more likely to form false memories compared to a control group.
• The false memories persisted over time, with 36.8% of participants holding onto inaccurate recollections even after one week.

Why it matters: The study highlights the potential risks of using AI in sensitive areas like legal interviews or psychological therapy, where memory reliability is crucial. As AI continues to play a larger role in human interactions, these findings emphasize the need for ethical guidelines and regulatory frameworks to prevent unintended cognitive distortions. In contexts where accuracy and truth are paramount, mitigating the risk of AI-induced false memories will be vital for maintaining trust in AI systems.

Trending AI Tools

• QWiser - Transform courses into interactive study material
• Storydoc - Create interactive slide decks to enhance engagement
• Amoihub - Manage generative AI art prompts with ease
• Intuit Assist - Intelligent GenAI financial insights for small businesses
• Jamboss - Turn your ideas into full-length songs
• Maximum - Automated solutions for software development

QUICK HITS

Anthropic launched a Quickstarts repository to help developers build applications with their API, starting with a Claude-powered customer support agent app.

Google announced the release of Function Calling in Google AI Studio, allowing users to test model capabilities directly within the user interface.

Ethereum founder Vitalik Buterin suggested that using AI and prediction markets could accelerate X’s Community Notes feature for quicker fact-checking.

The FDA approved Rebrain’s OptimMRI software update, which includes a new AI model to help target brain regions for treating Parkinson’s disease.

Dutch regulators fined Clearview AI more than $30 million for GDPR violations, including building a database of biometric data without consent.

Ligo released an open-source implementation of DeepMind’s AlphaFold3 protein structure prediction model.

A new benchmark of 80+ large language models found Anthropic’s Sonnet 3.5 performed best overall, while top models varied significantly by programming language.

Thank you for reading our newsletter! If you want to stay two steps ahead of the competition, subscribe to this newsletter. If you want to leave your competition in the past, hop on a quick, complimentary, no-obligation call with our team to explore our consulting and custom development services.

We've proudly worked with over 400+ companies to revolutionize their business with AI, and our team of 4,000+ developers, engineers, consultants, and experts are more than ready to help you take advantage of all the latest and greatest AI technology for your business.

Ready to get started? Book a Consultation today!