Singular Intel #16
API customers and DreamStudio users can now access the beta version of Stable Diffusion XL.
Stability AI has released Stabile Diffusion XL (SDXL), an image generation model that excels in photorealism and is aimed at enterprise clients. SDXL produces more detailed imagery with enhanced composition and face generation capabilities. It also has the ability to generate legible text and offers functionality beyond text-to-image prompting, including image-to-image prompting, inpainting, and outpainting. SDXL powers Stability AI's premium consumer imaging application DreamStudio and popular third-party apps like NightCafe Studio. The open-source version of SDXL will be released soon.
Amazon has now entered the generative AI race with the introduction of Bedrock.
Amazon has launched Amazon Bedrock, a service that allows customers to build generative AI-powered apps via pretrained models from startups, including AI21 Labs, Anthropic and Stability AI. The models hosted on Bedrock include multilingual text-generating models and a suite of text-to-image Bedrock-hosted models. AWS customers can also customise any Bedrock model by pointing the service at a few labelled examples in Amazon S3 cloud storage. The move is Amazon's most forceful play yet for the generative AI market, which could be worth close to $110bn by 2030, according to estimates from Grand View Research.
Alibaba, a tech giant from China, is set to release a rival to ChatGPT that utilizes artificial intelligence.
Alibaba's cloud computing unit has announced plans to release its own generative AI chatbot, called Tongyi Qianwen, which will be integrated into the company's businesses. The chatbot will initially be added to Alibaba's workplace messaging app and will perform tasks such as writing emails and drafting business proposals. The technology is capable of working in English as well as Chinese. Other technology companies including Microsoft, Google and Baidu have also released their own AI models and chatbots. However, concerns have been raised over the potential risks of powerful AI systems, with some calling for training to be suspended. Recently, Italy became the first Western nation to block ChatGPT due to privacy concerns.
According to reports, Elon Musk has bought numerous GPUs for a generative AI project on Twitter.
Despite calling for a pause in the development of powerful AI models, Twitter CEO Elon Musk has reportedly purchased around 10,000 GPUs for a generative AI project within Twitter. The exact nature of the project is unknown, but it is believed to be a large language model (LLM) that could potentially utilize Twitter's user tweets to train for natural language output. The purchase of GPUs represents a significant commitment towards developing deep-learning AI models within Twitter and could be used to improve Twitter search or create novel advertising methods. The investment in GPUs is estimated to be in the tens of millions of dollars and will likely be housed in Twitter's Atlanta data center. Some have criticized Musk's calls for caution on the subject of AI development, claiming that his recent moves undermine his previous statements.
Adobe has released a new video editing tool that uses AI to enhance text-based editing.
Adobe has announced updates to its family of video applications, including an AI-powered text-based video editing feature for Premiere Pro to make life simpler for professionals and creators. The company also celebrated 30 years of After Effects, adding a context-sensitive Properties Panel and community-requested improvements. Adobe's latest innovations are designed to address the workflow needs of today's professional editors and motion designers by automating time-intensive tasks, freeing creative professionals to spend more time shaping the stories they want to tell. Additionally, Adobe has expanded Frame.io's industry-leading video collaboration platform to photography and PDF documents. Premiere Pro is the only professional editing software to incorporate text-based editing, which makes video editing as simple as copying and pasting text.
An AI tool has been developed to detect early signs of lung cancer that may go unnoticed by doctors.
Researchers in Massachusetts have developed an AI-powered tool called Sybil that can detect signs of lung cancer years before doctors would be able to. The tool analyzes a 3D image of the patient's lungs and seeks out growths and patterns that may indicate cancer. Sybil has an accuracy rate ranging from 75% to 94%, according to studies conducted using thousands of volunteers' CT data. The tool holds promise as a potentially life-saving tool, but the team acknowledges the lack of diversity in its training data and hopes to diversify it before seeking FDA approval.
Infogrid has secured $90 million in funding for its technology that utilizes AI to monitor buildings.
Infogrid, a startup that uses AI to collect and analyze data on air quality, occupancy, and energy consumption, has raised $90 million in a Series B round led by Northzone. The funding will be used for product development, strategic hires, and customer acquisition efforts. Infogrid's platform gathers data from IoT sensors to help property managers optimize the performance of their buildings. The company competes with other building management systems but offers more context-specific data such as occupancy and air quality. The pandemic has driven interest in remote monitoring and real estate consolidation trends.
Prominent experts are urging Europe to broaden its A.I. regulations.
A group of more than 50 AI experts and institutions are calling on European lawmakers to broaden their regulations of AI, including general purpose AI (GPAI), rather than just focusing on high-risk AI. The group recommends that regulation should be considered around how AI is developed, including how data is collected and who is involved in training the technology. They also suggest that the legislation should be future-proofed and not restricted to specific types of products, like chatbots. The group warns against developers avoiding liability through standard legal disclaimers. This broad regulation could become a global precedent for future AI legislation.
OpenAI’s Bug Bounty Program
OpenAI has launched a Bug Bounty Program to encourage security researchers, ethical hackers, and technology enthusiasts to identify and address vulnerabilities in their AI systems. Through this program, individuals can report vulnerabilities, bugs or security flaws they discover in OpenAI's systems. The company has partnered with Bugcrowd, a leading bug bounty platform, to manage submissions and rewards. OpenAI will offer cash rewards ranging from $200 to $20,000 based on the severity and impact of reported issues. OpenAI views security as a collaborative effort and invites the security research community to participate in their Bug Bounty Program.
HyperWrite has revealed a revolutionary AI agent that is capable of browsing the internet in a way that closely resembles human behavior.
HyperWrite, a startup that specializes in AI writing, has unveiled its personal assistant AI agent that can interact with websites like a human user. Although the capability to enter credit card information will only be enabled for select test users at first, the agent is intended to handle basic web tasks on command. However, experts have warned that such software could be prone to human vulnerabilities and misused for malicious activities like phishing or hacking. HyperWrite is focused on deploying AI agents safely and ensuring that users have control over their actions. To try the personal assistant AI agent, install the Chrome extension and request early access.
Alphabet has invested $100 million in AlphaSense, a startup focused on AI.
Alphabet's CapitalG has led a $100 million funding round in AlphaSense, valuing the company at $1.8 billion. AlphaSense competes with financial data companies such as Bloomberg and FactSet and will use the funds to incorporate large language models into its products, including a feature that summarises financial documents for customers. AlphaSense CEO Jack Kokko said the money would also be used to improve technology and hire additional salespeople as the company prepares to go public. The latest investment follows months of increased hype surrounding generative AI, particularly OpenAI's ChatGPT and other text-generating tools that use large language models.
Elon Musk founds new AI company called X.AI
Elon Musk has started a new company called X.AI dedicated to artificial intelligence, according to Nevada filings. Musk is currently the director and Jared Birchall, the director of Musk's family office, is listed as its secretary. Musk reportedly sought funding from SpaceX and Tesla investors to get the company started. The company's name matches the branding of the X Corp. name he has since assigned to Twitter, along with the "X" label he's applied to his vision of an "everything app." Musk has been openly opposed to OpenAI, a rival AI organization that he co-founded in 2015 but walked away from in 2018, and recently signed a letter calling for a pause on "giant AI experiments."
Meta AI releases AI ‘Animated Drawings’
Meta has released an annotated dataset of nearly 180,000 amateur drawings to help AI researchers and creators innovate further. The dataset is unique as it reflects real-world conditions, including blurriness, hard shadows and crinkled surfaces. An AI system research demo was also created to bring artwork to life through animation. The demo allowed people to upload images and receive a short animation of their humanlike character within their drawing. The system incorporates repurposed computer vision models trained on photographs of real-world objects, which were fine-tuned with the Amateur Drawings Dataset. The code and dataset are being shared in hope that it will inspire a new generation of creators with its expressive and accessible possibilities.
Dolly is a ChatGPT-style AI model that is both free and open source.
Databricks has released Dolly 2.0, an open-source, instruction-following large language model (LLM) that has been fine-tuned on a human-generated data set. Dolly can be used for commercial purposes without paying for API access or sharing data with third parties. The model is based on EleutherAI's pythia model family and is exclusively fine-tuned on training data crowdsourced from Databricks employees. The resulting data set and Dolly's model weights and training code have been released fully open source under a Creative Commons license, enabling anyone to use them for any purpose, including commercial applications. Dolly 2.0 could potentially spark a new wave of open-source language models that are not hampered by proprietary limitations or restrictions on commercial use.
Deloitte has launched a new practice aimed at assisting companies in implementing generative AI.
Consulting firm Deloitte Touche Tohmatsu has established a new practice to help enterprises take advantage of generative artificial intelligence (AI), the technology behind chatbots and assistants. Deloitte's Generative AI Market Incubator will help companies implement third-party generative AI solutions and build their own, while a research and development team will train and fine-tune foundation models. The new practice will work with Deloitte's AI Academy to train people in skills such as model development and prompt engineering. Areas where generative AI can be applied include developer productivity, fraud detection, supply chain optimisation and smart factories.