What’s New in AI?

By Brandon Gottshall

Here's a summary of recent developments in artificial intelligence, including advancements in AI models, applications, and industry trends.

Meta's Llama 3 Release

Meta released Llama 3, its latest open AI model, on April 18, 2024. Key points include:

• Two initial models with 8 billion and 70 billion parameters.

• Trained on 15 trillion tokens of data, significantly more than Llama 2.

• Improved performance on benchmarks like MMLU without increasing model size.

• Open license for research and commercial use.

• Plans for a 400 billion parameter model later in 2024.

The Llama 3 400B Model

400B means 400 billion parameters. GPT-4 is speculated to have 200-300 billion parameters. While technically less efficient in terms of computational resources, the model is open source and can be run on anyone's equipment without an additional licensing cost burden.

The 400B model is particularly noteworthy:

• Expected to achieve near-parity with GPT-4 on benchmarks

• Could democratize access to state-of-the-art AI capabilities

• Potentially rivaling the performance of GPT-4 and Claude Opus at lower computational costs

Sources:

IEEE Spectrum

Toms Guide

Advancements in AI Models and Technologies

• OpenAI's Project Strawberry (previously Q-Star) reportedly progresses on AI systems with enhanced reasoning capabilities.

  • Source: Tom's Guide

• Researchers have developed an innovative camera mechanism inspired by the human eye to enhance robot vision.

  • Source: Real Clear Science

• Meta plans to expand Llama 3's capabilities:

  • Multimodality and multilingual support

  • Longer context windows

  • Source: Meta

AI in Business and Industry

• Samsung experienced a substantial operating profit increase of 932.8% for the first quarter of 2024 compared to the first quarter of 2023. The AI industry and chip demand drove the rise.

  • Source: CNBC

• Samsung anticipates a substantial profit increase, exceeding 1,400%, driven by the AI industry and chip demand.

  • Source: Electronics Specifier

• Enterprises are exploring AI systems for workforce management, potentially replacing human managers.

  • Source: BBC

• Microsoft and Nvidia have demonstrated foresight in recognizing AI opportunities, outpacing competitors like Apple. Microsoft and Nvidia demonstrated this foresight through their significant revenue growth in the AI sector. It does look like Apple may be able to still catch up with their unique offering Apple Intelligence, which seems to be somewhat of a platform for AI Providers. This careful planning and execution, and the unique ecosystem may greatly increase the odds of Apple having quick growth in the space.

  • Sources:

    • Nvidia News

    • The Guardian

    • IT Pro

AI Applications and Innovations

• Researchers are developing AI tools to interpret dog vocalizations. These tools are sort of like sentiment analysis for canines. They can tell whether your dog's bark is playful or aggressive with an accuracy of 70%.

  • Source: University of Michigan - Alumni Association

• A new fictional writing tool called Novel AI allows users to generate well-defined narratives, allowing them to define the world, the lore, and the characters to build complex stories.

AI in Finance and Cryptocurrencies

Market Trends

The cryptocurrency market exhibited strong bullish sentiment for AI-related tokens in early July. Investors are increasingly optimistic about integrating AI technologies in the financial sector, driving up the value of these tokens.

• Source: CoinDesk

Concerns and Challenges

• Security Risks: There are growing concerns about the potential hacking, misuse, and unintentional biases of AI systems within the financial sector and politics, including malicious or manipulative information campaigns. The DHS has even warned of the potential threats posed to the upcoming election by AI-generated misinformation intended to confuse or overwhelm voters. These risks necessitate robust security measures and ongoing vigilance.

  • Sources:

    • Reuters

    • Reuters

    • ABC News

    • PYMNTS

• Bias Reduction: Researchers have developed a cost-effective training method to reduce social biases in AI systems. This advancement ensures fairness and equity in AI applications across various industries, including finance.

  • Oregon State University

These developments showcase the rapid progress in AI across various sectors, from language models and robotics to business applications and environmental monitoring. The advancements also highlight ongoing challenges in AI development, including security concerns and the need to address potential biases.

Stay tuned for more updates!