Is Open Source Dying?
Stability AI appoints new CEO, Anthropic's Claude 3.5 Sonnet outperforms GPT-4o, OpenAI's DALL·E 3 enhances text-to-image AI, showcasing AI's rapid progress across industries.
In March 2024, Emad Mostaque, the CEO and co-founder of Stability AI, announced his resignation from the company. Mostaque is leaving Stability AI to focus on decentralizing AI and ensuring it benefits humanity. He believes that centralized AI controlled by a few companies or governments could lead to a loss of freedom and potentially harm humanity. Mostaque's vision is that decentralized, collective intelligence is the only way to counter this threat.
While it is good news for a new space. Decentralized AI. It’s wasn’t necessarily an exciting news for the open source community.
We wrote in our this issue, how this move was not only inspired by a “new” space, rather also by Stability AI’s lack of financial path to profitability, while open source is cool, reality is companies who open source esp large ones, struggle to make money.
Which is probably also why Emad moved on.
Since then there was an interim CEO, we didn’t however see a necessarily "open source” release for Stable Diffusion 3. It was released under a non-commercial license, which beats the whole “open source” story because you can’t now sell the model by making tweaks.
Last week, Stability AI appointed Prem Akkaraju, the former CEO of visual effects company Weta Digital, as its new CEO. It seems though he’s mostly going to be aligned with making profits to help company survive.
Gut says open source esp. at Stability AI is dying today.
It’s a shame though after putting out SOTA models like SDXL, the progress has slowed down.
Anyho, for curious minds, What is Decentralized AI?
Decentralized AI uses blockchain to distribute data and processing across a network, avoiding centralized control and promoting transparency, security, and accessibility.
There’s a bunch of different updates today as well:
Anthropic's latest language model, Claude 3.5 Sonnet, has outperformed GPT-4o and other competitors across various benchmarks, showcasing advancements in areas like coding, reasoning, and visual understanding.
Additionally, OpenAI has released DALL·E 3, the latest version of its groundbreaking text-to-image AI model. DALL·E 3 boasts an enhanced understanding of nuance and detail, allowing users to more easily translate complex ideas and descriptions into precise, high-quality images.
Let’s dive in!
Stuff you should know
Stability AI’s New CEO
British AI startup Stability AI has appointed Prem Akkaraju, the former CEO of visual effects company Weta Digital, as its new CEO, according to a report by The Information.Akkaraju is part of a group of investors, including former Facebook President Sean Parker, that is providing a cash infusion to the cash-strapped Stability AI. This investment from Akkaraju's investor group could result in a lower valuation for the AI company, as it has been seeking a large equity investment to fund its ambitious plans to challenge tech giants like OpenAI and Google in the AI race.
This move comes as Stability AI navigates a competitive market and seeks to solidify its position as a leading force in the development of transformative AI technologies. The cash infusion from Akkaraju's investor group could provide the necessary resources for the company to accelerate its product roadmap and better compete with the deep-pocketed tech giants dominating the AI field.
Claude 3.5 Sonnet beats GPT-4o
Claude 3.5 Sonnet is Anthropic's latest large language model, representing a significant advancement over previous versions with marked improvements in grasping nuance, humor, and complex instructions, allowing it to outperform competitor models like GPT-4o, Gemini-1.5 Pro, and Llama-400b across a range of benchmarks including coding proficiency, graduate-level reasoning, and undergraduate-level knowledge, while operating at twice the speed of the previous Claude 3 Opus model at a more cost-effective rate, making it well-suited for tasks like context-sensitive customer support and multi-step workflows, and demonstrating strong visual reasoning skills that surpass its predecessor in interpreting charts/graphs and accurately transcribing text from imperfect images, enhancing its value for industries like retail, logistics, and finance that rely heavily on insights from visual data, alongside the launch of a new feature called "Artefacts" that allows users to generate and collaborate on content like code snippets, website designs, and documents within the Claud.AI platform.
DALLE-3 gets an upgrade!
DALL·E 3 is the latest version of OpenAI's groundbreaking text-to-image AI model, representing a significant advancement in the company's efforts to bridge the gap between language and visual representation. This model boasts an enhanced understanding of nuance and detail, allowing users to more easily translate complex ideas and descriptions into precise, high-quality images. Notably, DALL·E 3 is built natively on OpenAI's language model, ChatGPT, enabling users to leverage the conversational AI as a brainstorming partner to refine prompts and generate tailored, detailed instructions to guide the image generation process. Recognizing the potential for misuse, DALL·E 3 incorporates improved safety mitigations, including the ability to decline requests for images of public figures by name and address harmful biases related to visual over- or under-representation. OpenAI is also researching ways to help identify when an image was created using DALL·E 3, such as a provenance classifier, to increase transparency and trust in the use of AI-generated imagery.
Hand Picked Video
Anthropic's Claude 3.5 Sonnet sets new benchmarks in reasoning, knowledge, and coding, outperforming competitors and previous models with improved intelligence, speed, and mid-tier pricing
Top AI Products from this week
Tellers.ai - Turn any text into video edits using real world footage. Tellers is not generating fake clips instead it leverages any database of videos to find the most relevant clips to illustrate your text.
Cheerleader AI - Get the ultimate cheerleading experience with Cheerleader AI! 🎉 Boost your motivation and productivity with your very own virtual cheerleader!
Licode Page for PreSales and Waitlist - Augie is a video creation and editing studio that enables businesses to leverage the power of video-first marketing.
Make Podcast - MakePodcast.io allows you to create professional podcasts in minutes using AI voices. Simply provide your script, select your voices, and let our AI produce a professional-quality podcast in minutes!
BuildSpace Sage - sage is a new place on the internet for builders, makers, & creators to find and be found by the right group of ppl.
Meeno - Meeno provides everyday relationship advice - friends, family, romance, workplace. Informed by experts, powered by AI.
KnoWhiz - KnoWhiz is an AI-powered learning tool to revolutionize autonomous learning. Users input learning objectives or textual materials, and KnoWhiz can auto-generate structured learning, and adjust learning content from cumulative outcomes like quizzes and tests.
BloxGPT - Streamline your workflow with AI-powered chats on a collaborative whiteboard, giving you the flexibility to take your productivity to the next level.
This week in AI
Microsoft drops Florence-2 - Microsoft has released Florence-2, a new unified vision foundation model that can handle a variety of tasks like captioning, object detection, and segmentation, performing on par or better than large vision models. Florence-2 is available in 232M and 771M parameter sizes and is expected to provide enterprises a single, versatile approach to vision applications, reducing the need for separate task-specific models.
DALLE -3 Upgraded Image Generator Model - OpenAI has upgraded DALL·E to version 3, a significant advancement in text-to-image AI that leverages ChatGPT for enhanced nuance and detail. DALL·E 3 incorporates safety mitigations, provenance tools, and will be available to ChatGPT Plus, Team and Enterprise customers in October, allowing users to utilize generated images without permission.
AI in finance - Citigroup's report forecasts AI will drive $2T in banking profits by 2028, but raises concerns over data security, regulation, and agile firms outpacing incumbents in AI adoption.
Apple-Meta AI Partnership Talks - Apple and Meta are in talks to integrate Meta's generative AI into Apple Intelligence, as Apple seeks AI partners to enhance its capabilities and offer premium subscription options through its platform.