Mastercard has instructed financial institutions to cease facilitating marijuana transactions via its debit cards, affecting an industry already on the periphery of the US financial system. This is due to the fact that despite several states legalizing medicinal and recreational marijuana, it remains federally illegal. As such, most banks in the US do not service cannabis companies. A Mastercard spokesperson stated that they had instructed institutions offering payment services to cannabis merchants to end these activities because cannabis sales are considered illegal by the federal government. Industry leaders like Sunburn Cannabis CEO Brady Cobb and Verano’s President Darren Weiss expressed concern over this development, stating it would further hinder the state-legal cannabis industry and impede consumer access. The decision comes amid uncertainty over the future of legislation such as the SAFE Banking Act, which aims to ease access to banking services for the cannabis industry but has been criticized by Republican Senator John Cornyn as “wishful thinking”. (Reuters)
The increase in usage of e-bikes, particularly in New York City, has led to a surge in fires and deaths caused by their lithium-ion batteries. The rise in incidents has instigated calls for regulation of how these batteries are manufactured, sold, charged, and stored. Advocates and fire departments are urging the U.S. Consumer Product Safety Commission to create national safety standards and seize imports that do not comply. The issue will be discussed in a public hearing in Washington. New York has experienced 100 battery-related fires so far this year, with 13 deaths. New York’s Senators, Chuck Schumer and Kirsten Gillibrand, introduced legislation to establish safety standards for e-bikes and their batteries. The city also enacted local laws, including a ban on sale or rental of uncertified e-bikes and batteries, and received a $25 million federal grant for e-bike charging stations to reduce fire risks. Battery-related fires are on the rise globally, with incidents reported in London and San Francisco. (Associated Press)
Meta has released Llama 2, an open-source and free-to-use version of its large language model, posing a competitive challenge to proprietary systems like OpenAI’s ChatGPT. Llama 2, created with 40% more data than its predecessor, claims to generate results on par with ChatGPT. It offers a cheaper alternative for developers and startups to create custom applications, potentially fuelling the AI boom. The model is available for use through partnerships with companies such as Hugging Face, Databricks, and OctoML, and is also supported by Microsoft and Amazon’s AWS. Ahmad Al-Dahle, Meta’s VP for generative AI, highlighted the company’s commitment to open-source, citing demand beyond the research community. Despite this, the model’s licensing agreement restricts companies with over 700 million monthly active users, and the details of the model’s training data remain undisclosed. The release of Llama 2 has been likened to Google’s launch of the Android OS, providing a powerful yet affordable alternative that could stimulate innovation. Nathan Lambert, an AI researcher at Hugging Face, anticipates later versions of Llama catching up to proprietary models. The shift towards open-source models, however, comes with concerns over legal liability and potential misuse. (Wired)
Meta has introduced Llama 2, a large language model (LLM) with up to 70 billion parameters, that’s claimed to be on par with OpenAI’s GPT 3.5 in academic benchmarks. Microsoft and Qualcomm have announced support and partnership for Llama 2. Notably, Qualcomm plans to bring Llama 2 to smartphones by 2024. Llama 2’s models are free to use for commercial and academic purposes and compact enough to run on less powerful devices. However, a license clause requires companies with over 700 million monthly active users to seek separate permission to use Llama 2.
Llama 2 also applies a technique called Ghost Attention (GAtt) that aids in dialogue control over multiple turns. This enhancement, along with the model’s diverse range and strong performance relative to size, increases its applicability. The model’s open-source license is seen as an accelerator, allowing developers to tweak the model for specific needs, particularly for devices like smartphones. Despite some outperforming models on the Open LLM leaderboard, they are based on Meta’s original Llama model, suggesting that Llama 2 is poised to ascend the charts as developers refine it. (IEEE Spectrum)
Human-made noise from ships and oil and gas infrastructure in the Arctic Ocean can disrupt narwhal behavior, causing these marine animals to make shallower dives and cut their hunts short, according to a study published in Science Advances. Scientists, conservationists, and Inuit communities, who rely on narwhals as a food source, are concerned about how development and ship traffic will impact the whales. The study focused on the effects of noise pollution associated with oil exploration activities. The researchers, who used tracking tags to monitor narwhal reactions to boat noises, found that as soon as a ship was within hearing distance, the narwhals aborted their deep dives and ceased their underwater sonar clicking. The researchers recommend establishing shipping routes that avoid narwhal habitats to keep their foraging grounds quiet and protect them in the increasingly noisy Arctic waters. (Science)
Thanks for reading. Let’s be careful out there.
WORDS: The Biology Guy.
IMAGE CREDIT: NASA.