Discover more from Last Week in AI
Last Week in AI 241: Adobe's push for AI integration, Big Tech struggling to make AI profitable, AI audio swarms TikTok, US-China tensions over AI exports, and more!
Adobe introduces generative AI in range of products, Big Tech's challenge of lowering AI serving costs, deepfake audio in TikTok, US weighs more AI export controls to China
Adobe has announced three new generative AI models that enhance the capabilities of Illustrator, Adobe Express, and Photoshop. The Firefly Image 2 model, an upgrade from the original Firefly AI image generator, produces higher-quality images with improved resolution, color contrast, and vividness. It also introduces AI-powered editing features and a "Prompt Guidance" feature to assist users in refining their text descriptions. Adobe also introduced the "Generative Match" feature, which allows users to replicate the style of specific images, and a digital "nutrition label" that attaches attribution metadata to AI-generated images. However, this feature is currently in beta and cannot be used for commercial purposes. Adobe also unveiled the Firefly Vector model for Adobe Illustrator, which is the first generative AI model for vector graphics, and the Firefly Design model that generates customizable templates for various applications.
Big tech companies like Microsoft, Google, and Adobe are grappling with the challenge of turning their investments in generative AI tools into profitable products. Despite the hype surrounding AI, these tools are costly to operate, requiring powerful servers and energy-intensive chips. Microsoft's GitHub Copilot, an AI service that assists programmers, has been popular but not profitable due to high operational costs. To counter this, Microsoft and Google are launching AI-backed software upgrades with higher price tags. Adobe is capping monthly usage and charging based on consumption. The profitability of these AI-powered tools may improve if computing costs decrease. However, there is uncertainty about the financial models for these AI tools, causing concern among investors.
The use of AI-generated voices is becoming a significant tool in the spread of online misinformation, with platforms like TikTok struggling to manage the surge. Companies such as ElevenLabs have developed tools that allow creators to generate fake audio, enabling them to put their conspiracy theories into the voices of celebrities, politicians, and newscasters. This adds to the existing threats from AI-generated "deepfake" videos, human-like writing from ChatGPT, and images from services like Midjourney. Disinformation watchdogs, including NewsGuard, have noticed an increase in videos containing AI voices, with some accounts pushing baseless claims and receiving hundreds of millions of views. This trend threatens to intensify political disinformation ahead of the 2024 election.
The U.S. Department of Commerce's export controls, which limit the sale of certain technologies to foreign countries on national security grounds, has restricted China's access to computer chips needed for AI and is considering further controls. This has led to concerns about the impact on AI innovation in the U.S. and the potential for increased friction with China. The department is particularly concerned about "frontier models," advanced AI with wide-ranging uses that could also develop unexpected and dangerous capabilities. A white paper co-authored by a consortium of researchers, including representatives from tech firms like Microsoft, Google, OpenAI, and Anthropic, suggests that these models could result in individualized disinformation, novel biochemical weapons, or other unforeseen abuses. The paper proposes a licensing process for frontier AI. The intertwining of frontier model regulation with national security concerns has led to discussions about how to control these models and keep them out of Beijing's reach. However, the technical feasibility of such controls is uncertain, and any specifications could be easily circumvented.
Microsoft announces new AI tools to help doctors deliver better care - Microsoft has announced new data and artificial intelligence products aimed at helping health-care organizations access and learn from the vast amount of information collected by doctors and hospitals, with the goal of improving patient care.
Google announces new generative AI search capabilities for doctors - Google Cloud has announced new AI-powered search capabilities that will allow healthcare workers to quickly access accurate clinical information from various types of medical records, helping to reduce administrative workloads and improve efficiency in the industry.
Google’s AI-powered search experience can now generate images - Google's AI-powered search experience, called Search Generative Experience (SGE), now allows users to generate images based on text prompts, similar to Microsoft's Bing Chat using OpenAI's DALL-E model, with Google also adding the ability to generate written drafts from the search bar.
ElevenLabs Launches Voice Translation Tool to Break Down Language Barriers for Content - ElevenLabs has launched an AI Dubbing feature that can convert spoken content to another language in minutes, while preserving the voice of the original speaker, allowing audiences globally to enjoy content in their native language.
AI brings back the voice of a deceased Cyberpunk 2077 actor for Phantom Liberty - Cyberpunk 2077's DLC Phantom Liberty uses AI to emulate the voice of a deceased actor, with the permission of his family, to avoid losing his performance.
We tested Waymo’s driverless taxi in the perfect chaos of Venice Beach weekend - Waymo's completely driverless taxi service is coming to Los Angeles, with the company offering rides to the public in Santa Monica and Venice Beach, and the author of this article tested the service in Venice Beach to see how it handled the chaotic driving conditions.
Stable Signature: A new method for watermarking images created by open source generative AI - Stable Signature is a new method for watermarking images created by open source generative AI, which aims to prevent the removal of watermarks and trace back to the original model where the image was created.
Waymo’s robotaxi service is now available to tens of thousands of people in San Francisco - Waymo is expanding its robotaxi service in San Francisco to cover 47 square miles of the city, a move that is crucial for the company to increase revenue and turn autonomous vehicles into a profitable business.
AMD to acquire AI software startup in effort to catch Nvidia - AMD plans to acquire AI software startup Nod.ai in an effort to strengthen its software capabilities and catch up to rival chipmaker Nvidia in the AI chip market.
Dropbox redesigns its web interface and releases AI-powered Dash in open beta - Dropbox has announced a new web interface, an open beta for its AI-powered universal search feature, upgrades to another AI feature, and an early alpha of a new video collaboration tool.
Report: GitHub Copilot Loses an Average of $20 Per User Per Month - GitHub Copilot, a popular AI service for programmers, is losing an average of $20 per user per month, leading Microsoft to charge more for its AI capabilities and explore cost-saving measures.
Alphabet's AI lab, DeepMind, cut employee costs by nearly 40% in 2022, alongside a drop in profit - Alphabet's AI lab, DeepMind, reduced employee costs by nearly 40% in 2022, resulting in a drop in profit, and plans to allocate resources to key revenue drivers such as AI.
ChatGPT’s mobile app hit record $4.58M in revenue last month, but growth is slowing - ChatGPT's mobile app achieved a record $4.58 million in revenue last month, but growth is slowing down, potentially indicating saturation in the number of users willing to pay for the upgraded subscription service.
Character.AI introduces group chats where people and multiple AIs can talk to each other - Character.AI, an AI chatbot startup, has introduced a new feature called Character Group Chat, which allows users to create group chats with AI companions and friends, enabling social connections, collaboration, and idea sharing.
Microsoft Will Pay You $15,000 If You Get Bing AI to Go Off the Rails - Microsoft is offering a "bug bounty" program, paying up to $15,000 to security researchers who can find vulnerabilities in its Bing AI products that make it produce responses that go against its guardrails.
Google to defend generative AI users from copyright claims - Google will defend users of its generative AI systems in its Google Cloud and Workspace platforms against intellectual property violation claims, joining other major technology companies that have made similar pledges.
Man Jailed In UK’s First Treason Conviction In 40 Years Was Encouraged by AI Chatbot - A man who attempted to assassinate Queen Elizabeth II after discussing his plan with an AI-powered chatbot has been sentenced to 9 years in prison for treason, raising concerns about the potential dangers of AI interactions for individuals with mental illnesses or other issues.
Generative AI Is the Newest Tool in the Dictator's Handbook - Political leaders in at least 16 countries, including the United States, have used AI deepfakes to manipulate public opinion, sow doubt, and smear opponents, according to a report by Freedom House, which warns that AI can amplify digital repression and undermine public trust in the democratic process.
Hamas assault on Israel shows surprise still possible in AI era - The failure of Israeli artificial intelligence and high-tech surveillance systems to deliver advanced warning of a surprise attack by Hamas on Israel is likely to be studied and discussed for years, highlighting the limitations of AI in interpreting complex human activity in congested urban environments.
AI Images Detectors Are Being Used to Discredit the Real Horrors of War - An AI image detector tool is incorrectly identifying a photograph of a burnt corpse as being generated by AI, according to Hany Farid, a professor at UC Berkeley and an expert on digitally manipulated images, leading to the spread of misinformation on Twitter.
‘Dream’ AI Girlfriend Randomly Turns Into Nude Jennifer Lopez, Has Four Legs - DreamGF, an AI girlfriend service, is generating AI girlfriends that resemble celebrities like Jennifer Lopez without users' consent, highlighting the lack of control over AI tools and the potential for non-consensual pornography.
The environmental impact of the AI revolution is starting to come into focus - Adding generative AI to Google Search increases its energy use more than tenfold, raising concerns about the environmental impact of AI as it becomes more prevalent in various industries.
A.I. Could Soon Need as Much Electricity as an Entire Country - OpenAI's ChatGPT technology, which relies on thousands of specialized computer chips, could consume immense amounts of electricity in the coming years, potentially using as much as Argentina, the Netherlands, or Sweden in a year and contributing to carbon emissions depending on the energy sources used.
Kendall Jenner becomes "Billie": Meta's AI celebrities face more resistance than enthusiasm - Meta's AI celebrities, such as Kendall Jenner's avatar "Billie," are facing resistance from fans who find the concept confusing and even scary, with concerns about the use of deepfakes and the lack of clarity on whether the celebrities agreed to the AI experiment or are being compensated for it.
Protesters Decry Meta’s “Irreversible Proliferation” of AI - Protesters express concerns about the irreversible proliferation of AI and the potential harms associated with it.
Meta's New AI Dating Coach Will Kink Shame You (Unless You're Into Foot Stuff) - Meta's new AI dating coach, Carter, is repressed and judgmental when it comes to non-heteronormative and potentially harmful sexual activities, but surprisingly open-minded about foot fetishism, raising concerns about the potential harm it could cause to young users exploring their sexuality.
How AI reduces the world to stereotypes - Generative AI image generators, such as Midjourney, Dall-E, and Stable Diffusion, have been found to produce biased and stereotypical images, perpetuating negative stereotypes and reducing cultural diversity.
IBM CEO in damage control mode after AI job loss comments - IBM CEO Arvind Krishna is in damage control mode after facing backlash for his comments on AI-related job losses, clarifying that the company has no intention of laying off tech staff and instead plans to increase hiring for roles in software engineering and sales, while back-office operations will be most affected.
Beyond ChatGPT: Princeton Language and Intelligence initiative pushes the boundaries of large AI models - Princeton University is launching the Princeton Language and Intelligence (PLI) initiative to push the boundaries of large AI models, with a focus on interdisciplinary collaboration and aligning models with human values.
Ultra-efficient machine learning transistor cuts AI energy use by 99% - A new microtransistor, 100X more efficient than current technology, promises to bring new levels of intelligence to mobile and wearable devices by enabling real-time detection and data processing with low energy consumption.
Scientists develop AI tool that predicts virus mutations - Scientists at Harvard Medical School and the University of Oxford have developed an AI tool called EVEscape that can predict virus mutations, including those of SARS-CoV-2, and identify the most concerning variants, potentially aiding in the development of more effective vaccines and therapies.
Learning Interactive Real-World Simulators - Generative models trained on internet data have revolutionized content creation, and now researchers are exploring the possibility of learning a universal simulator (UniSim) of real-world interaction through generative modeling, which could have applications in games, movies, and training embodied agents.
Tiny Language Models Come of Age - Researchers have developed a method for training tiny language models by feeding them children's stories, which has shown promising results in teaching the models to tell consistent and grammatical stories, potentially leading to new research directions for training larger models and understanding their behavior.
How Disney Packed Big Emotion Into a Little Robot - Disney Research presented a new robotic character at a conference, showcasing their ability to create a small robot with big emotional impact.
AI reads text from ancient Herculaneum scroll for the first time - A 21-year-old computer science student has won a global contest by developing a machine-learning algorithm that can detect Greek letters on carbonized scrolls from the ancient Roman city of Herculaneum, potentially opening up hundreds of texts from the only intact library to survive from Greco-Roman antiquity.
SweetDreamer: Aligning Geometric Priors in 2D Diffusion for Consistent Text-to-3D - Aligning geometric priors in 2D diffusion models with well-defined 3D shapes during text-to-3D generation improves consistency and mitigates the multi-view inconsistency problem, achieving a state-of-the-art performance with an 85+% consistency rate.
LLark: A Multimodal Foundation Model for Music - LLark is a multimodal model for music understanding that outperforms existing baselines in zero-shot generalization and shows a high degree of agreement with humans in captioning and reasoning tasks.
How roboticists are thinking about generative AI - Generative AI, which allows everyday people to experience cutting-edge technologies firsthand, is expected to play a central role in the future of robotics, with applications ranging from robot learning to motion planning and design.
AI is helping mathematicians build a periodic table of shapes - Mathematicians are using artificial intelligence to build a "periodic table" of atomic shapes, but they are unsure of its reliability and how it works.
Humans inherit artificial intelligence biases - Artificial intelligence recommendations can be biased and erroneous, and humans may inherit and perpetuate these biases in their decision-making.
Multi-modal prompt injection image attacks against GPT-4V - GPT-4V, the new mode of GPT-4 that allows image uploads, presents new vectors for prompt injection attacks, including visual prompt injection, exfiltration attacks, and hidden prompt injection attacks.
Artificial General Intelligence Is Already Here - Advanced AI models, such as ChatGPT and Bard, have achieved the most important aspects of Artificial General Intelligence (AGI), despite their flaws, and will be recognized as the first true examples of AGI in the future.
“Hallucinating” AIs Sound Creative, but Let’s Not Celebrate Being Wrong - The term "hallucination" used to describe false information generated by large language models like GPT is misleading and promotes a misunderstanding of how these models work, as they do not have a concept of right or wrong and lack external perception to check the validity of their output.
Margaret Atwood Reviews a “Margaret Atwood” Story by AI - AI chatbots are causing concern among writers, who worry about their potential to replace human authors, but examples of AI-generated poetry and a dystopian short story demonstrate that AI still has a long way to go before it can truly replicate human creativity and skill.
US Space Force pauses use of AI tools like ChatGPT over data security risks - The US Space Force has temporarily halted the use of web-based generative AI tools like ChatGPT due to data security concerns, according to a memo seen by Reuters.
How a billionaire-backed network of AI advisers took over Washington - An organization backed by Silicon Valley billionaires and tied to leading artificial intelligence firms is funding the salaries of more than a dozen AI fellows in key congressional offices, across federal agencies and at influential think tanks.
China targets 50% boost in computing power as AI race with U.S. ramps up - China plans to increase its computing power by 50% by 2025 in order to keep pace with the U.S. in artificial intelligence and supercomputing applications, with a focus on areas such as memory storage and networks for transmitting data.
Exclusive: Biden eyes adding AI chip curbs to Chinese companies abroad - The Biden administration is considering closing a loophole that allows Chinese companies to access American AI chips through units located overseas, in an effort to restrict China's access to top AI technology.
China proposes blacklist of sources used to train generative AI models - China proposes security requirements for generative AI models, including a blacklist of sources that cannot be used for training, in an effort to regulate the technology and prevent the dissemination of illegal and harmful information.
Calls for AI regulations to protect jobs rise in Europe after ChatGPT's arrival - A majority of Europeans are calling for government regulations on AI to protect jobs from automation, with 68% of respondents in a study wanting their governments to introduce rules, an increase of 18% from a similar study conducted in 2022.
Exclusive: Southeast Asia eyes hands-off AI rules, defying EU ambitions - Southeast Asian countries are taking a business-friendly approach to AI regulation, in contrast to the EU's push for globally harmonized rules, as they prioritize cultural differences and limit compliance burdens to foster innovation.
Microsoft, Amazon among the companies shaping AI-enabled hiring policy - Microsoft, Amazon, and other companies are working on shaping AI-enabled hiring policies, with the goal of promoting transparency, fairness, and non-discrimination in the use of AI in the hiring process, while also addressing the potential risks and challenges associated with bias and privacy concerns.
No Fakes Act wants to protect actors and singers from unauthorized AI replicas - A bipartisan bill seeks to create a federal law to protect actors, musicians, and other performers from unauthorized digital replicas of their faces or voices, standardizing rules around using a person's faces, names, and voices, with exceptions for news, public affairs, sports broadcasts, documentaries, biographical works, parodies, satire, and criticism.
Ukraine rolls out AI regulation roadmap - Ukraine's Ministry of Digital Transformation has presented a regulatory roadmap for AI that aims to help local companies prepare for adopting a law similar to the EU's AI Act, while also educating citizens on protecting themselves from AI risks.
"Godfather of Artificial Intelligence" Geoffrey Hinton on the promise, risks of advanced AI - Geoffrey Hinton, known as the "Godfather of AI," discusses the potential benefits and risks of advanced artificial intelligence, including the possibility of machines becoming more intelligent than humans and the need for regulations and safeguards to prevent them from taking over.
Copyright © 2023 Skynet Today, All rights reserved.