Discover more from Last Week in AI
Last Week in AI #189: State of AI 2022, AI-generated clip art, reducing waste with ML, and more!
This year's State of AI Report details the latest in AI research, industry, politics, and safety, Microsoft to deploy text-to-image models in consumer apps, Google releases models for waste management
This year’s State of AI Report was just released - it outlines important developments across AI research, industry, politics and safety. Research saw the rise of diffusion models, community-driven open-sourcing of large models, and a shifting of scaling to focus on bigger datasets in addition to bigger models. Industry saw the continued dominance of Nvidia’s chips, faster research to commercialization cycles, and a swath of new startups founded by prominent researchers amid tech company hiring freezes. For politics, when it comes to large models, it is clear that all of the work is being done in industry, and none in academia, and new research labs outside of academia and funded by non-traditional sources are more and more common. For AI safety, more and more in the field are expressing concern, but research activity in AI safety is nowhere close to those on the capabilities side.
Our take: The State of AI Report has been measured and insightful over the years. We recommend readers to take a skim - there are many interesting details, and the report’s predictions are always fun to ponder about.
Microsoft is building an AI text-to-image app, Microsoft Designer, that may eventually make it into its Office software. The company is only offering a limited web preview for now, but it plans to release it as a standalone app in the future. A similar app, called Image Creator, will be integrated into Bing. These are all similar to DALL-E 2, the text-to-image model that started it all built, by OpenAI, in which Microsoft invested a billion dollars back in 2019. The company also has an exclusive license in using OpenAI’s GPT-3, but it’s unclear whether similar licenses apply to DALL-E 2. There are many potential concerns about the ethics and legality of using large AI models trained on public image data for profit, and Microsoft’s push to commercialization may additional relevant discussions and debate.
Our take: If 2022 is the year where text-to-anything models came out of research labs, 2023 will be the year they make a big splash in commercial applications. Integration of such powerful generative models into creativity apps is the most straight forward application. Microsoft’s investment in OpenAI has proved to be worth their while, as the tech giant will not only have insider access to OpenAI’s latest research and models, it will also accumulate the know-how of training and hosting large models for ubiquitous consumer apps. It is no coincidence that Microsoft is the first FAANG company to turn generative models into consumer apps. It will be interesting to see who follows suit next and what new applications they can bring to the table.
Summary: Google introduced CircularNet, a set of object detection and segmentation models specialized for waste/recyclables detection. The team developed a customized and globally-applicable taxonomy of material types (e.g. paper, metal) and material forms (e.g. can, bottle) to handle the wide variety of recyclable goods. Data for the model came from Material Recovery Facilities, which provided large image datasets of different kinds of trash. Google has released the trained models as well as data preprocessing guidelines.
Our Take: By now it’s clear that requiring trash and recyclables to be cleanly sorted before processing by a waste management plant is too difficult. Building plants that are robust to cluttered and diverse material flows via automation is crucial for effective recycling and minimizing landfill use. It’s interesting that Google is releasing this model for free online instead of directly commercializing it in partnerships with waste management facilities. Perhaps the models are still not performant enough, or merely identifying material types is not sufficient to solve deliver significant economic value.
Growth in AI and robotics research accelerates - "By filtering across the disciplinary spectrum, the fields continue to hit new heights."
Should AI have a role in assessing research quality? - "A UK study aims to find out whether artificial intelligence could ease the peer-review process for the country’s Research Excellence Framework."
Google AI Introduces Frame Interpolation for Large Motion (FILM): A New Neural... - "Many studies are increasingly focusing on frame interpolation, which synthesizes intermediate pictures between a pair of input frames. The refresh rate can be increased, or slow-motion videos can be created using temporal up-sampling. "
Artificial intelligence could soon diagnose illness based on the sound of your voice - "Voices offer lots of information. Turns out, they can even help diagnose an illness — and researchers are working on an app for that. The National Institutes of Health is funding a massive research project to collect voice data and develop an AI that could diagnose people based on their speech."
AI Language Models Are Struggling to “Get” Math - "Should this be telling us something?"
UL2 20B: An Open Source Unified Language Learner - "In “Unifying Language Learning Paradigms”, we present a novel language pre-training paradigm called Unified Language Learner (UL2) that improves the performance of language models universally across datasets and setups."
AI scans Hurricane Ian damage to give cash to poorest victims - "Nonprofit GiveDirectly teamed up with Google to identify those most at need of financial support, using AI to cross reference aerial images of storm damage with poverty data in Florida and Puerto Rico."
DispatchTrack launches AI-powered CO2 tracking solution - "Last-mile delivery solutions provider DispatchTrack has launched an AI-powered carbon emissions tracking system to help companies meet their supply chain sustainability goals."
AI fire helmet expected to be the future of firefighting - "Scottish scientists designed an AI fire helmet that provides firefighters with real-time information about the fire."
This podcast brings Steve Jobs back to life, thanks to AI - "The episode does begin with a big disclaimer that the entire conversation has been generated by AI. "
Machine learning's potential for players and developers - "Haven Studios' Jade Raymond and Leon O'Reilly joined by Mark Cerny in discussing how new tech could change games, from conceptualization to customization"
Great, now the AI is coming for your grandma’s recipes as well! - "he Estonian startup Yummy started off creating a meal-kit startup, but along the way created an AI that can create and adapt recipes based on your taste and dietary restrictions, complete with AI-generated images of what your dishes might look like."
James Cook Hospital team cut out tumours aided by robots - "Surgeons using robots can now remove lung tumours with a single incision. Surgeons at Middlesbrough's James Cook University Hospital are said to be the first in the UK to perform the pioneering procedure."
Microsoft open-sources AI algorithms for optimizing farm operations - "Microsoft Corp. today open-sourced FarmVibes.AI, a collection of artificial intelligence models that farm operators can use to more efficiently perform tasks such as planting crops. "
Study looks at risk factors to predict hospitalization outcomes by developing machine learning model - "By using artificial intelligence, Houston Methodist researchers are able to predict hospitalization outcomes of geriatric patients with dementia on the first or second day of hospital admission."
AI Tool Can Detect Signs of Mental Health Decline in Text Messages - "University of Washington School of Medicine researchers have found that an artificial intelligence tool can accurately identify “red-flag language” in text messages from patients with serious mental illness."
Deep learning makes X-ray CT inspection of 3D-printed parts faster, more accurate - "A new deep-learning framework developed at the Department of Energy's Oak Ridge National Laboratory is speeding up the process of inspecting additively manufactured metal parts using X-ray computed tomography, or CT, while increasing the accuracy of the results."
Mortgage Fraud AI Improves Detection By 30% - "An AI document forensics solution improved fraud detection at an online mortgage broker by 30 percent, alongside reducing workload for investigators."
Stability AI, the startup behind the hot text-to-image art generator Stable Diffusion, quietly raised funding at a $1 billion valuation - "Stability AI, the creators of the hyper-popular text-to-image generator Stable Diffusion, has closed a new funding round that values the company at $1 billion, multiple sources familiar with the deal tell Insider. "
Meet Cassie, the Usain Bolt of robots - "Cassie holds the Guinness world record for fastest 100 meter dash for a bipedal robot, paving the way for robots to become more lifelike"
AI and Machine Learning Enter the Kitchen at Chipotle - "Chipotle’s next era of technology has begun to take shape inside restaurants, with a central focus on optimization. On Tuesday, the fast casual announced it’s piloting multiple technologies to streamline operations and strip friction from its business."
Xembly raises cash to develop an AI assistant for corporate meetings - "Macroeconomic factors are placing pressure on companies to improve efficiency, and on employees and their managers to do more with less."
Cybersecurity Will Account for Nearly One-Quarter of AI Software Market Through 2025 - "By 2025, the artificial intelligence (AI) software market will expand from 2021's $33 billion to $64 billion, according to a new report. And cybersecurity is the fastest-growing category of AI spend, experiencing a rise in spending of 22.3% compound annual growth rate (CAGR)."
Volkswagen to take 60% stake in $2 bln tech JV with China's Horizon Robotics - "Volkswagen's software unit Cariad will spend more than $2 billion and take a 60% stake in a new venture with Chinese technology firm Horizon Robotics, it said on Thursday, tackling two big challenges: software and the Chinese market."
Instagram expands AI-powered age verification program to India and Brazil - "Instagram, facing scrutiny from safety advocates, started testing a program in the U.S. earlier this year to verify users’ age claiming to be 18 or older. The program, uses techniques including authentication via running video selfies through an artificial intelligence system."
Claims AI can boost workplace diversity are 'spurious and dangerous' - "Recent years have seen the emergence of AI tools marketed as an answer to lack of diversity in the workforce, from use of chatbots and CV scrapers to line up prospective candidates, through to analysis software for video interviews."
Tech firms say laws to protect us from bad AI will limit ‘innovation’. Well, good - "For too long, the industry has escaped legal liability in the pursuit of its own interests – and the EU has had enough"
How AI and machine learning are changing the phishing game - "By combining AI and breached data, hackers are creating more targeted and sophisticated phishing campaigns and finding greater success."
Approach to Demystify Black Box AI Not Ready for Prime Time - "Research suggests that compared with human clinicians, image heat maps underperform and require further refinement"
The messy morality of letting AI make life-and-death decisions - "Automation can help us make hard choices, but it can’t do it alone."
Thief Steals Genshin Impact Fan Art Using AI, Demands Credit From Creator - "An AI art generator allowed a thief to ‘finish’ a digital painting before its creator could post it"
Sci-Fi Reddit Community Bans AI-Art for Being 'Low Effort' Posting - "Several artistic communities and platforms that host user-generated content have taken a stand against AI-generated artworks in recent weeks."
Data protection regulators publish myth-busting guidance on machine learning - "In its proposed AI Regulation (“AI Act”), the EU recognises AI as one of the most important technologies of the 21st century. It is often forgotten, however, that AI is not one specific type of technology."
A Day in the Life of a Machine Learning Engineer - "It is good to get a better insight into what other people’s day-to-day looks like. Many students are more focused on the skills, courses, and knowledge level they need to ensure they are as good as they can get. But sometimes, all you need is to hear it from the horse's mouth."
Copyright © 2022 Skynet Today, All rights reserved.