- AI KATANA
- Posts
- Meta launches Llama 2, an open source AI model that allows commercial applications
Meta launches Llama 2, an open source AI model that allows commercial applications
Also: OpenAI will give local news millions to experiment with AI
AI KATANA
July 19, 2023
Welcome!
Today, we delve into the exciting realm of open-source AI models, with Meta launching Llama 2, a family of models that have reportedly outperformed most other chat models in tests. We also look into OpenAI's generous initiative to fund local news in exploring the potential applications of AI within the sector. Meanwhile, generative AI is seen as a game-changer, potentially adding trillions in annual value to the global economy, while Microsoft continues to strengthen its AI offerings with a new subscription for enterprise users. We'll also share the latest AI tool updates, with image search being incorporated into Microsoft's Bing Chat and Google testing AI-generated backgrounds for its Meet video platform. Finally, we discuss the latest venture capital updates, where insurtech Tractable has raised $65M in a Series E funding round, and despite a dip in the VC landscape, AI is seen as a potential catalyst to turn things around. Dive in for all these stories and more!
Sliced:
🦙 Meta launches Llama 2, an open source AI model that allows commercial applications
📰 OpenAI will give local news millions to experiment with AI
🧑🏽💻 The great acceleration: CIO perspectives on generative AI
🧑🏽✈️ Microsoft Announces $30 AI Copilot Subscription
Meta has announced the release of Llama 2, an open source family of AI language models that offer commercial licenses and range in size from 7 to 70 billion parameters. Llama 2 has reportedly outperformed most other open source chat models in tests, according to Meta. Available on platforms such as Microsoft Azure, Llama 2's pretrained models are trained on 2 trillion tokens with a 4,096 token context window, while its fine-tuned models for chat applications have been trained on over 1 million human annotations. Despite controversy over the potential misuse of open source AI models, Meta maintains that such models foster transparency, economic competition, free speech, and democratic access to AI technology. Although Llama 2's performance can't match OpenAI's GPT-4, it is still considered a strong contender in the open source model field. Nevertheless, concerns about the lack of transparency regarding the training data for Llama 2 persist.
OpenAI is partnering with the American Journalism Project (AJP) and offering $5 million in funds to explore the potential uses of AI in supporting local news. Over a two-year period, the funds will be used to evaluate AI's potential within the sector, and to assist approximately ten of AJP's current grantees in developing their own AI tools. OpenAI will also give AJP up to $5 million in API credits to experiment with the technology. This move comes amidst the growing use of AI in various aspects of news reporting and the concerns it raises, including potential misinformation issues and the displacement of reporters' jobs. The collaboration aims to explore how AI could improve journalistic workflows, enabling more focus on in-depth reporting and sustainable operations.
Generative AI is poised to significantly impact global economies and enterprises, potentially adding between $2.6 and $4.4 trillion in annual value to the global economy, and accelerating the automation of work, according to McKinsey Global Institute and Goldman Sachs. Large language models (LLMs) such as ChatGPT are at the forefront of this evolution, offering unprecedented conversational capabilities. Although AI has long been recognized as strategically important, generative AI represents a key inflection point, with its potential to democratize AI across all functions of an enterprise. However, CIOs and technical leaders must make careful strategic decisions regarding data infrastructure, model ownership, workforce structure, and AI governance. Key considerations include leveraging generative AI to surface and utilize hidden data, building flexible, scalable, and efficient data infrastructures, potentially developing customized LLMs, addressing automation anxieties, and ensuring unified and consistent governance to manage risks and accelerate innovation.
Microsoft has announced a $30 per month AI subscription for enterprise software users named Microsoft Copilot, which is a premium variant of its 365 subscription. Unveiled in March, Microsoft Copilot leverages large language models to analyse data in Microsoft 365 applications. Alongside, the company has also launched an AI-powered tool called Bing Enterprise Chat aimed at business customers for applications such as real-time transcribing and copywriting. The $30 AI subscription offers businesses access to these advanced capabilities and more, as Microsoft continues to double down on AI partnerships.
🛠️ AI tools updates
Microsoft has announced a significant update to its Bing Chat AI bot, enabling users to conduct searches using images. The AI tool can now understand, interpret, and answer questions about uploaded images, a feature aligned with the ongoing advancements in generative AI by tech giants such as Microsoft, Google, OpenAI, and Anthropic. This update is part of a broader trend of incorporating image search into chatbot user experience, though none of the leading text-based chatbots currently generate their own images.
Google is currently testing AI-generated backgrounds for its Meet video platform, a feature spotted by tech enthusiast Artem Russakovskii. The functionality, being experimented within Google's Workspace Labs, enables users to describe their preferred backdrop, choose from various categories and apply vision effects during meetings. However, the feature is still in the rollout phase and may not be accessible to all users. Those interested can request to join the Workspace Labs testing program through Google's site. A crucial point to note is that Google's Workspace Labs privacy policy states human reviewers may read, annotate, and process your data to enhance the services.
💵 Venture Capital updates
London-based insurtech Tractable, known for leveraging AI for visual claims assessments and disaster recovery, has raised $65M in a Series E funding round led by SoftBank Vision Fund 2, with participation from existing investors Insight Partners and Georgian. Tractable intends to use these funds to boost its research and development capabilities and enhance the end-user experience with improved and integrated vehicle assessments. With a focus on digitising the laborious and expensive insurance claims process, Tractable's AI offers a ten-fold faster solution by analysing user-submitted photos of damaged vehicles or properties.
Venture Capital (VC) funding is facing a downturn, with the number of deals decreasing by 9% and their value declining by 34% in Q2, amidst economic and geopolitical uncertainties. Despite these issues, including inflation, high interest rates, supply-chain problems, and geopolitical tensions, there is optimism about stabilization driven by artificial intelligence, biotech, and energy sectors, according to EY's US VC leader, Jeff Grabow. Although many startups are still relying on funds obtained in 2020 and 2021, total VC funding is expected to exceed $100 billion in 2023 for the sixth consecutive year. While late-stage markets continue to struggle due to low IPO activity, seed-stage and Series A deals accounted for half of the VC deals, raising $7.2 billion. Reports suggest that despite a considerable drop in Q2, AI could spearhead a VC funding resurgence, keeping annual totals in line with 2020 figures.
🫡 Meme of the day
⭐️ Generative AI image of the day
Before you go, check out how Unilever Is Transforming Ice Cream With AI 🍦
