Last Week in AI #186: OpenAI's new speech-to-text AI, Nvidia's generative model for realistic 3D object, Ukrainian AI startup to produce the voice of Darth Vader, and more!
OpenAI's TTS model performs much better on noisy speech, Nvidia's GET3D generates 3D meshes while trained only on 2D images, James Earl Jones signed over rights to voice work for AI reproduction
On September 21, OpenAI open-sourced Whisper, an automatic speech recognition (ASR) system that enables "robust" transcription in multiple languages and translation from those languages to English. OpenAI wrote that Whisper is primarily targeted at AI researchers "studying robustness, generalization, capabilities, biases, and constraints of the current model," but that it can also be useful as an ASR solution for developers. The model performs much better than prior available systems on noisy real-world speech data, and the trained model is readily available on Github.
Our Take: Another cool technology from OpenAI--the question is what will be done with it? It appears that OpenAI is attempting to pursue the productization of models like DALL-E 2 and GPT-3 in parallel with more fundamental research. I think it's good that they are open-sourcing models like Whisper for the purpose of enabling academic study, but I also wonder if the insights and improvements gleaned from this will ultimately be used in research that is shared openly or in a gated product.
Nvidia recently released GET3D, a neural network model that can generate realistic 3D shapes with “high-fidelity textures and complex geometric details” that can be directly imported into 3D modeling software and game engines. Importantly, this model was trained entirely on 2D images, making data collection for this and similar methods in the future a lot more manageable (there is a lot more image data on the Internet than data of 3D models). GET3D is small enough to run on a single GPU, and the code will be uploaded next week.
Our take: The speed of advancement in generative AI is impressive as always, and what’s cool is that we’re seeing substantial and simultaneous developments across many modalities (text, images, sound, 3D models, videos, code) and that many of these methods are open-sourced. Non-trivial and economically valuable products can already be built from these off-the-shelf models and publically available APIs. It is a very exciting moment in AI and I can’t wait to see where the field goes, even just a few months from now.
James Earl Jones, who voiced the iconic Darth Vadar character for decades, is 91 this year and has expressed the desire to step away from this role in the past. To preserve his voice for the Star Wars franchise, the voice actor has signed over the rights to his previous work so that it can be used by AI voice acting companies for replication. Respeecher, a Ukrainian start-up, was hired by Disney to produce lines for Darth Vadar in the recent Obi-Wan Kenobi series (as well as the voice of the young Luke Skywalker in The Book of Boba Fett). Respeecher employees, based in Kyiv, persevered working on the series during Russia’s attacks on the capital. The co-founder of the company said in an interview how they live and work embody the resilience of Ukrainians, and he hopes that “more people will hear about Ukraine—about our tech community, about our start-ups—because of it.”
Our take: There are so many interesting angles to this story. From the way AI in creative applications is making a real impact in pop culture, to the implications of celebrities signing over their likeness so that they can be reproduced by AI, to the globalized nature of AI talent as well as the struggles and triumphs of the Ukrainian people and businesses in the last few months. From the AI perspective, a future where AI replicates the likeness of pop culture icons to lengthen and monetize existing franchises is already upon us. While this is great for profit margins and our collective sense of nostalgia, there is the concern that this will further reduce the number of original works and the opportunities for new stars to rise.
Can eyes on self-driving cars reduce accidents? Cues from moving eyes could help pedestrians anticipate vehicle's intentions - "Robotic eyes on autonomous vehicles could improve pedestrian safety, according to a new study at the University of Tokyo. "
Users trust AI as much as humans for flagging problematic content - "Social media users may trust artificial intelligence -- AI -- as much as human editors to flag hate speech and harmful content, according to researchers at Penn State."
New tool overcomes major hurdle in clinical AI design - "Harvard Medical School scientists and colleagues at Stanford University have developed an artificial intelligence diagnostic tool that can detect diseases on chest X-rays directly from natural-language descriptions contained in accompanying clinical reports."
AlphaFold developers win US $3-million Breakthrough Prize - "DeepMind’s system for predicting the 3D structure of proteins is among five recipients of science’s most lucrative awards."
Brains on board: Smart microrobots walk autonomously - "A collaborative effort has installed electronic “brains” on solar-powered robots that are 100 to 250 micrometers in size – smaller than an ant’s head – so that they can walk autonomously without being externally controlled."
NIH Grant Funds Machine Learning Tool for Chronic Disease Detection - "The nearly $3 million grant supports the development of a machine learning algorithm that intends to improve the detection of the chronic disease, pulmonary hypertension."
A swarm of 3D printing drones for construction and repair - "An international research team led by drone expert Mirko Kovac of Empa and Imperial College London has taken bees as a model to develop a swarm of cooperative, 3D-printing drones."
AI predicts lung cancer tumor growth after radiation : NHS study - "The resulting AI model was able to predict a patient’s two-year risk of recurrence more accurately than current methods, including the widely used TNM staging system."
NGA Puts Machine Learning to Work to Speed Mission, Further Research - "The National Geospatial-Intelligence Agency is well known for analysis of imagery and maps, but text, or written language, is a key part of the process."
Why DeepMind isn’t deploying its new AI chatbot — and what it means for responsible AI - "DeepMind considers Sparrow a research-based, proof-of-concept model that is not ready to be deployed"
T-Mobile’s 5G fights fire with Pano AI detection system - "T-Mobile announced that it has teamed up with Pano AI, a San Francisco-based disaster preparedness company, to deploy 5G-connected cameras for detecting active wildfires."
I Resurrected "Ugly Sonic" with Stable Diffusion Textual Inversion - "So there’s a new popular AI image generation tool named Stable Diffusion. But first, let’s discuss why you really clicked on the link to this article: Ugly Sonic."
This US political site has started using AI to illustrate its articles - "AI art generators are going mainstream"
How ad agencies are using AI image generators—and how they could be used in the future - "Ad Age asked more than 20 ad experts how DALL-E 2, Midjourney and Stable Diffusion are changing creative"
Instacart’s grocery ‘smart cart’ shows how it’s about more than just deliveries - "The Instacart platform could make your local grocery store a connected one"
Auto supplier Magna bringing electric delivery robots to Metro Detroit - "One of the robots has been working for a Detroit-area pizzeria since March, the company said."
Intel and Mila join forces for responsible artificial intelligence - "Intel yesterday announced a three-year strategic research and co-innovation collaboration with Mila, a Montreal-based artificial intelligence (AI) research institute."
China’s Factories Accelerate Robotics Push as Workforce Shrinks - "China installed almost as many robots in its factories last year as the rest of the world, accelerating a rush to automate and consolidate its manufacturing dominance even as its working-age population shrinks."
DeepMind Says It Had Nothing to Do With Research Paper Saying AI Could End Humanity - "The paper was published recently in the peer-reviewed AI Magazine, and was co-authored by researchers at Oxford University and by Marcus Hutter, an AI researcher who works at DeepMind."
CytoReason, Pfizer ink $110M, 5-year extension of AI-powered drug development deal- - "Three years after the pair first teamed up, they’ve signed on to continue their AI-fueled collaboration for at least another five years."
Tech war: Nvidia CEO Jensen Huang confident in China market despite US AI chip ban - "Nvidia founder and CEO Jensen Huang Jen-hsun said China continues to present a significant growth opportunity, even after the US government restricted the Californian company from exporting two of its top chips to the country."
Tesla sends out invites for AI Day 2, teases Full Self-Driving, Tesla Bot, Dojo, and more - "Tesla has started sending out invites for its AI Day 2022 and the automaker is teasing more information on Full Self-Driving, Tesla Bot, Dojo, and more."
Microsoft’s New Security Chief Looks to AI to Fight Hackers: Q&A - "Charlie Bell says new tools will help analyze data faster and alleviate shortage of security staff"
Hellblade Developer Ninja Theory confirms it won’t replace voice actors with AI - "Ninja Theory, the developer behind the Hellblade series, denies that it will replace human voice actors with AI."
Getty Images bans AI-generated content over fears of legal challenges - "Getty Images is worried about future copyright claims"
Clearview AI, Used by Police to Find Criminals, Now in Public Defenders’ Hands - "After a Florida man was accused of vehicular homicide, his lawyer used Clearview AI’s facial recognition software to prove his innocence. But other defense lawyers say Clearview’s offer rings hollow. "
AI Skills Crisis May Lead to Wasted Investments and Stifled Innovation: SAS - "How to Solve the Data Science Skills Shortage is a report based on a survey of decision makers from major US firms spanning nine sectors, including banking, insurance, government and retail."
Just how good are the new wave of AI image generation tools? - "Examining DALL-E, Midjourney and Stable Diffusion - and the implications for games development."
Of God and Machines - "The future of artificial intelligence is neither utopian nor dystopian—it’s something much more interesting."
There’s no driving test for self-driving cars in the US — but there should be - "Would a European-style regulatory system improve safety?"
Artist receives first known US copyright registration for generative AI art - "In what might be a first, a New York-based artist named Kris Kashtanova has received US copyright registration on their graphic novel that features AI-generated artwork"
Europe edges closer to a ban on facial recognition - "Should the EU ban software that can pick a face out of a crowd?"
Should the EU Regulate General-Purpose AI Systems? - "The French presidency recently proposed expanding the scope of the EU’s draft law for regulating high-risk AI tools to include a new class of systems: General-purpose AI systems, which are those that can perform a wide range of tasks and often power more specific applications."
Is AI Really A Job Killer? These Experts Say No - "Yes, AI and intelligent technology will take over some jobs, but that will free up workers to do more challenging and important work."
Meta's AI guru LeCun: Most of today's AI approaches will never lead to true intelligence - "I think AI systems need to be able to reason," says Yann LeCun, Meta's chief AI scientist. Today's popular AI approaches such as Transformers, many of which build upon his own pioneering work in the field, will not be sufficient."
Computer Scientist Explains One Concept in 5 Levels of Difficulty | WIRED - "Moravec's paradox is the observation that many things that are difficult to do for robots to do come easily to humans, and vice versa. Stanford University professor Chelsea Finn has been tasked to explain this concept to 5 different people; a child, a teen, a college student, a grad student, and an"
Copyright © 2022 Skynet Today, All rights reserved.