The Falcon Flies, Microsoft Will Defend You Against Frivolous AI Lawsuits and Claude Goes Pro

The Falcon Flies, Microsoft Will Defend You Against Frivolous AI Lawsuits and Claude Goes Pro

Unleashing Falcon 180B: A New Giant in AI Language Models

In the landscape of artificial intelligence, the arrival of Falcon 180B is akin to spotting a rare bird in the wild. The big open source model is a new heavyweight champ on the block, coming to us from the Technology Innovation Institute potentially knocking out the Meta mega-model Llama 2 that released only last month.?

It clocks in at an impressive 180 billion parameters making it the largest openly available language model to date. It doesn't just stand tall, it soars high, achieving state-of-the-art results and edging close to the performance of the top proprietary models. The Falcon 180B was trained on a colossal dataset, and its use, albeit under certain restrictive conditions, is a commercial possibility.

However, it's important to note that every bird has its quirks, and Falcon 180B is no exception. The base model, while powerful isn't instruct-tuned for conversational responses and it’s licensing model is restricted for commercial hosting, requiring a special license.

Once fine-tuning communities get their hands on it we’ll really see what it can do.? It’s marketing post says it’s the top model on the HuggingFace LLM leaderboard but it’s really not.? It’s the top foundation LLM there but it’s surpassed by fine-tunes of Llama 2.? That’s seems to the constant cycle of the moment.? A big company releases a new state of the art model and then the community gets to work fine-tuning it to beat the benchmarks.? Time will tell if fine-tuning communities can deliver something more than a benchmark beater.? We’d love to see the community evolved into one that’s trying to add new knowledge and better responses to a much wider range of possible use cases.? Time will tell if it’s any more than a hobby and mushrooms into a movement the way we see open source teams fixing bugs and adding new capabilities to traditional code.

Microsoft's Copilot Commitment: A Copyright Safe Haven for Users

In a bid to address the intricate challenges of copyright in the AI era, 微软 has introduced a new initiative - the Copilot Copyright Commitment. This move assures customers that they can freely use Copilot services and its generated output without fretting over potential copyright claims. Microsoft is stepping up, promising to take on any legal risks tied to copyright challenges, even to the point of defending and covering costs for any adverse judgments or settlements resulting from copyright infringement lawsuits. The tech giant is also implementing filters and safety systems to minimize the risk of infringing content, all while respecting creators' rights and not claiming any intellectual property rights in the outputs of Copilot services.

Ideogram's Bold Leap into AI-Generated Imagery

In the world of AI, Ideogram is making waves as a generative image startup, having successfully raised a cool $16.5 million in seed funding. Their mission? To revolutionize creativity with AI, focusing on the challenge of reliable text generation in AI-produced images. They're turning heads with a variety of preset image generation styles from typography to fashion, and their web app, still in beta, is already getting positive reviews for its lettering and image generation. They may be missing a few features that rival image generators offer and have had a minor hiccup rendering their own name, but their unique offering of a reliable typographic generator is winning over graphic designers. Competition is heating up, though, with other AI image generators like Midjourney also upping their game.

Meet Persimmon-8B: The New Open-Source Language Model on the Block

Step aside LLaMA2 and GPT-3, there's a new language model in town. Adept has generously open-sourced Persimmon-8B, a language model that's got less than 10 billion parameters but packs a punch with a context size of 16K. Despite being trained on less data, it's giving the likes of LLaMA2 a run for their money. The model is a decoder-only transformer with modifications, and it's got some nifty features. It comes with unused embeddings for multimodal extensions, sparse activations, and a combination of C++ speed and Python flexibility. While it doesn't include strategies for controlling toxic outputs, it's still a powerhouse.

Claude Pro: Anthropic's New Premium Plan for ChatGPT Rival

In the world of AI chatbots, competition is heating up as Anthropic rolls out Claude Pro, a premium subscription plan for its ChatGPT competitor. For just $20 a month, the same as ChatGPT Plus, users will enjoy enhanced features such as faster response times, priority access to new features, and increased model performance. Not only does Claude Pro offer up to five times more usage than the free model, Claude 2, but it also includes longer conversations and more messaging.

Also this week:

要查看或添加评论,请登录

AI Infrastructure Alliance的更多文章

社区洞察

其他会员也浏览了