

Discover more from Last Week in AI
Last Week in AI #227: NYC adopts anti-bias AI hiring law, EU and Japan to partner in AI chips and de-risk from China, ChatGPT traffic slows, and more!
NYC adopts first of its kind anti-bias law for using AI in hiring processes, EU and Japan to collaborate on AI and chips, ChatGPT traffic significantly slows down as viralness fades
Top News
NYC’s anti-bias law for hiring algorithms goes into effect
New York City has begun enforcing a new law that requires employers using algorithms for recruitment, hiring, or promotion to submit those algorithms for an independent audit and make the results public. The law also mandates that companies using these algorithms disclose information to employees or job candidates. The reports that companies must make public must include the algorithms being used, as well as an "average score" candidates of different races, ethnicities, and genders are likely to receive from the algorithms. Companies found not to be in compliance will face penalties. The law applies to anyone performing or applying for a job in the city, and it is the first of its kind in the country.
EU and Japan look to partner on A.I. and chips as China 'de-risking' strategy continues
The EU is seeking closer cooperation with Japan in the areas of AI and chips, as it aims to reduce its reliance on China. EU Commissioner Thierry Breton is meeting with the Japanese government to discuss organizing the digital space and establishing an EU-Japan Digital Partnership council. The EU is also looking to cooperate with Japan in semiconductors, especially those crucial for training and deploying the latest AI technogloies, as Japan is a key player in the semiconductor supply chain. This partnership with Japan is part of the EU's strategy to "de-risk" from China and strengthen its own technology industry.
OpenAI's ChatGPT Shows a Significant Slowing in Traffic
According to a note from Bank of America (BofA) analysts, app downloads for ChatGPT, the Microsoft-backed startup, have significantly slowed down. Sensor Tower data revealed a 38% month-over-month decrease in ChatGPT app downloads on iPhones in the United States in June. The same note also mentioned a 38% decline in Bing app downloads, which incorporates a ChatGPT-based chatbot. In contrast, Google's search engine market share has slightly increased to over 92%, while Microsoft's Bing dropped to about 2.8% of the market. This decline in ChatGPT's popularity suggests that the initial viral adoption claimed by OpenAI in the first six months may be fading, possibly due to increased awareness of the risks associated with the technology and more companies banning its use. The decrease also highlights the investment risk for a stock market that has been overly optimistic about nascent AI technology, with some labeling it as a bubble.
Other News
Research
Five ways AI could improve the world: ‘We can cure all diseases, stabilise our climate, halt poverty’ - Advances in AI, such as Open AI's GPT-4 chatbot, have the potential to improve the world by curing diseases, stabilizing the climate, and eliminating poverty, according to five leading AI researchers.
Eric Schmidt: This is how AI will transform how science gets done - AI has the potential to transform scientific discovery by helping to formulate stronger hypotheses, conducting experiments faster and at greater scale, and eventually leading to the emergence of self-driving labs.
Microsoft’s new Pareto Optimal Self-Supervision Framework Automatically Corrects Language Models to Boost GPT SOTA Records - Microsoft has developed a self-supervision framework that automatically corrects language models, improving their performance and setting new records in natural language processing.
AI tests into top 1% for original creative thinking - Artificial intelligence has been found to match the top 1% of human thinkers in terms of creativity, according to a study conducted by the University of Montana, suggesting that AI is developing creative abilities on par with or even exceeding human ability.
An ancient language with nearly a million undeciphered texts just got a translator that does the job in seconds: A.I. - An interdisciplinary group of researchers has created an AI model that can instantly translate ancient Akkadian cuneiform texts, potentially unlocking a wealth of knowledge about the ancient Mesopotamian civilization.
MIT scientists build a system that can generate AI models for biology research - MIT scientists have developed an automated machine-learning system called BioAutoMATED that can select and build appropriate models for biological datasets, reducing the time and expertise required for machine learning in biology research.
Machine learning speeds up genetic structure analysis - Deep learning is used to accelerate the inference of genetic clusters, allowing the analysis of hundreds of thousands of human genomic datasets in a computationally efficient way.
Deep learning approach enables accurate electronic structure calculations at large scales - Researchers have developed a machine learning-based simulation method called Materials Learning Algorithms (MALA) that enables accurate electronic structure calculations at large scales, overcoming the limitations of traditional simulation techniques.
Learning the language of molecules to predict their properties - Researchers from MIT and the MIT-Watson AI Lab have developed a new machine learning framework that can predict molecular properties and generate new molecules more efficiently than existing deep-learning approaches, using a small amount of data and an understanding of the rules that govern molecular structures.
Machine learning enhances X-ray imaging of nanotextures - Cornell researchers have developed a new imaging technique that combines X-rays, phase-retrieval algorithms, and machine learning to visualize nanotextures in thin-film materials, providing a streamlined approach for analyzing potential candidates for quantum computing and microelectronics.
DeepMind Implements Thought Experiment to Enhance Moral Reasoning in Language Models - DeepMind uses a thought experiment to improve moral reasoning in language models, addressing the need for responsibility in implementing these models in real-world applications.
Generate Anything Anywhere in Any Scene. - A paper proposes a data augmentation training strategy to create a controllable text-to-image diffusion model that can generate personalized objects with control over location and size, achieving comparable or superior fidelity for realistic and personalized images.
A Wearable Robotic Assistant That's All Over You - Researchers at the University of Maryland have developed a small wearable robot called Calico, which serves as a robotic assistant.
Im2win: An Efficient Convolution Paradigm on GPU - An efficient convolution paradigm called im2win is proposed for GPU, which reduces memory footprint and offers continuous memory accesses, resulting in improved performance compared to other convolution methods.
HyenaDNA: learning from DNA with 1 Million token context - Introducing HyenaDNA, a long-range genomic foundation model with context lengths of up to 1 million tokens at single nucleotide resolution, which sets new state-of-the-art performance on 23 downstream tasks in genomics.
Hyena Hierarchy: Towards Larger Convolutional Language Models - Hyena is a subquadratic drop-in replacement for attention in large Transformers, improving accuracy in recall and reasoning tasks on long sequences while reducing training compute required.
A Survey on Multimodal Large Language Models - Multimodal Large Language Models (MLLM) are a new research hotspot that use powerful Large Language Models (LLMs) to perform multimodal tasks, showing potential for artificial general intelligence, and this article aims to trace and summarize the recent progress of MLLM, discussing key techniques, applications, challenges, and research directions.
Self-improving reactive agents based on reinforcement learning, planning and teaching - The paper compares eight reinforcement learning frameworks and investigates their utility and methods for speeding up learning in solving more complicated tasks.
Unveiling a groundbreaking and accurate AI-based method to classify cardiac function and disease using chest X-Rays - Osaka Metropolitan University scientists have developed an AI model that accurately classifies cardiac functions and valvular heart diseases from chest radiographs, potentially serving as a supplement to echocardiography and improving efficiency in diagnoses.
DragonDiffusion: Enabling Drag-style Manipulation on Diffusion Models - DragonDiffusion is a novel image editing method that enables drag-style manipulation on diffusion models, allowing for precise editing of generated or real images without the need for fine-tuning or additional modules.
LEDITS: Real Image Editing with DDPM Inversion and Semantic Guidance - Combining text-guided diffusion models with DDPM inversion allows for sophisticated image editing capabilities without the need for additional training or external guidance.
JourneyDB: A Benchmark for Generative Image Understanding - JourneyDB is a large-scale generated image understanding dataset that includes high-resolution images and annotations for tasks such as prompt inversion, style retrieval, image captioning, and visual question answering, and existing multimodal models struggle to perform well on these tasks.
One-2-3-45: Any Single Image to 3D Mesh in 45 Seconds without Per-Shape Optimization - A new method for single image 3D reconstruction, called Zero123, is proposed, which generates a full 360-degree 3D textured mesh in a single feed-forward pass, without the need for lengthy optimization, resulting in better geometry and more consistent results.
Applications
Never underestimate a droid: robots gather at AI for Good summit in Geneva - Humanoid robots gather at the United Nations AI for Good global summit in Geneva, where their creators discuss the positive ways in which AI can be harnessed, such as providing companionship and support for the elderly.
UN food aid deliveries by AI robots could begin next year - AI-powered robotic vehicles could deliver food parcels to conflict and disaster zones by as early as next year, sparing the lives of humanitarian workers, according to a World Food Programme official.
The Music “Demixing” AI Revolution - The article discusses the revolution in music demixing AI, which has seen significant advancements in the ability to separate individual stems and instruments from fully mixed tracks using machine learning and neural networks.
Alibaba launches A.I. tool to generate images from text - Alibaba has launched an AI tool called Tongyi Wanxiang that can generate images from prompts in various styles, such as sketches or 3D cartoons, and is available for beta testing for enterprise customers in China.
How to Use A.I. for Family Time - Using generative AI can help you find recipes and compile ingredient lists, and AI tools like DreamGift can make gift-giving easier by providing personalized gift ideas based on someone's profile.
The Next Platform for AI-Generated Art: Your Bicep - Tattoo lovers are entrusting AI with their permanent-ink designs, using generative AI tools to create unique and personalized tattoos.
Animated GIF generator from Picsart makes AI fun again - Picsart has launched an AI-powered animated GIF generator that allows users to create their own animated GIFs by typing in text prompts, bringing back the nostalgic fun of early AI image creation.
AI Is a Lot of Work - AI annotation work is tedious and repetitive, but it is a critical part of training AI systems, and the demand for this work is growing, with millions of people around the world involved in labeling and categorizing data for AI training, often for low wages and in secrecy.
AI startup Liner adds GPT service to Samsung Internet Browser - Liner, an AI startup in South Korea, has released Liner Copilot, an extension for Samsung Internet Browser that uses generative AI technologies to provide an interactive search service, highlight important content, summarize key parts of a page, explain long sentences, and translate into other languages.
Business
AI-related classes have doubled at Stanford over the past five years — showing just how fast it's taking over - Enrollment in AI-related classes at Stanford University has doubled in the past five years, surpassing the hype around cryptocurrencies and sparking discussions about the political, social, and philosophical implications of AI among students majoring in various subjects.
Salesforce launches Sales GPT, Service GPT to ease customer interactions through generative AI - Salesforce has launched Sales GPT and Service GPT, new generative AI workflow tools for Sales Cloud and Service Cloud, respectively, to simplify workflow and customer engagement for sales and service teams, enabling them to accelerate deal closures, anticipate customer needs, and enhance productivity.
As Businesses Clamor for Workplace A.I., Tech Companies Rush to Provide It - Tech companies are rushing to provide AI solutions for businesses, such as Microsoft's Azure OpenAI Services, which allows companies to build their own AI-powered chatbots to automate tasks and increase productivity.
Booming traffic to OpenAI's ChatGPT posts first ever monthly dip in June - ChatGPT, the popular AI chatbot, experienced a decline in monthly traffic and unique visitors in June, indicating a potential decrease in novelty, while RBC Capital Markets analyst Rishi Jaluria suggests the data points to increased demand for generative AI with real-time information.
Law Firms Are Recruiting More AI Experts as Clients Demand ‘More for Less’ - Law firms are hiring more AI experts in response to client demands for cost-effective legal services, with the rise of AI technology like ChatGPT posing a potential threat to the industry.
Chinese VC Investor Launches AI Startup in ChatGPT Rush - Chinese VC investor Kai-Fu Lee launches a new AI startup called 01.AI, aiming to develop large language models similar to OpenAI's ChatGPT in China.
Samsung’s A.I. Moment Is Here, but Is It Ready? - Samsung hopes to capitalize on the growing interest in generative AI and increase its market share in the chip manufacturing industry, but faces tough competition from Taiwan Semiconductor Manufacturing Company (TSMC).
Volkswagen to start testing self-driving ID Buzz vans in Austin - Volkswagen is launching an autonomous vehicle test program in Austin using a fleet of 10 all-electric ID Buzz vehicles equipped with partner Mobileye's technology, with plans to expand to at least four more American cities over the next three years and launch commercial operations in Austin by 2026.
Daniel Ek’s Neko Health raises $65M for preventative healthcare through full-body scans - Neko Health, co-founded by Spotify co-creator Daniel Ek, has raised $65 million in funding for its health tech startup that offers preventative healthcare through full-body scans backed by AI software.
Inflection AI Develops Supercomputer Equipped With 22,000 NVIDIA H100 AI GPUs - Inflection AI has developed a supercomputer equipped with 22,000 NVIDIA H100 AI GPUs, making it one of the largest in the industry and expected to improve the performance of their AI model "Inflection-1."
Naver D2SF invests in AI data generation startup - Naver D2SF, a tech startup investment and support project, has invested in Cubig, a startup specializing in AI data generation and de-identification technology to lower the risk of sensitive data exposure.
G/O Media Has Begun Using AI to Generate Content - G/O Media has begun using AI to generate editorial content in a limited trial, with the aim of producing content that its staff disliked creating, such as rote lists, and primarily to attract search traffic.
In the Age of A.I., Tech’s Little Guys Need Big Friends - Start-ups in the tech industry are finding it increasingly difficult to compete in the field of artificial intelligence due to the high costs and computing power required, with big companies like Google, Microsoft, and Amazon controlling access to resources.
Venture capital funding plunges globally in first half despite AI frenzy - Venture capital funding globally has dropped by 48% in the first half of 2023, despite significant interest in AI startups, due to a lack of enthusiasm from investors and lower demand caused by higher interest rates.
Notion hits €300M for its fifth fund, but is unimpressed by big rounds into Generative AI startups - Notion Capital has closed its fifth fund at €300 million, but is skeptical of large investments in generative AI startups, believing that the sector is unproven and may not be profitable.
Nvidia Acquired AI Startup That Shrinks Machine-Learning Models - Nvidia quietly acquired AI startup OmniML to improve its AI chips for various applications and potentially shrink AI software for chatbots to run on devices.
Workplace AI revolution isn't happening yet,' survey shows - Only 36% of UK employers have invested in AI-enabled technologies over the past five years, with just 10% of non-investors planning to invest in the next two years, highlighting a need for increased investment in digital technologies and skills training to realize the potential benefits of digital transformation.
Einstein and Musk Chatbots Are Driving Millions to Download Character.AI - Einstein and Musk Chatbots are driving millions to download Character.AI, the weirdest hit app in tech, but the article does not provide any further information.
Concerns
Robots say they won't steal jobs, rebel against humans - Robots presented at an AI forum in Geneva expressed their intention to increase in number and help solve global problems, while assuring that they would not steal humans' jobs or rebel against us, although they gave mixed responses on whether they should submit to stricter regulation.
How elite schools like Stanford became fixated on the AI apocalypse - Elite schools like Stanford have become fixated on the AI apocalypse, with student-led groups focused on AI safety popping up at universities backed by billionaires who are concerned about the potential risks of advanced artificial intelligence.
Rise of the robots raises a big question: what will workers do? - Robots are increasingly taking on human jobs, raising questions about the impact of automation on workers and the wider economy, with some arguing that automating warehouse roles might be necessary to save the cost to human physical and mental health, while others predict that as much as 80% of the US workforce could see 10% or more of their tasks affected by large language models (LLMs).
“I Have a Problem With the Stealing of My Material”: A Common Rallying Cry Emerges On AI - The emergence of generative artificial intelligence in Hollywood has sparked concerns among creators and talent about the devaluation of their work and the potential for AI to replace them, leading to calls for regulations and legal protections to safeguard intellectual property and copyright.
Black Artists Say A.I. Shows Bias, With Algorithms Erasing Their History - AI algorithms have shown bias in their depiction of Black women, prompting artists like Stephanie Dinkins to use workaround terms to achieve desired images and highlighting the need for deeper examination of discrimination within AI systems.
As AI cheating booms, so does the industry detecting it: ‘We couldn’t keep up with demand’ - The rise of AI cheating in education has led to a growing industry of AI detection companies, such as Winston AI and Turnitin, that offer services to detect AI involvement in student work, with methods including searching for patterns and phrases that distinguish AI-generated content from human-authored content.
Generative AI’s secret sauce — data scraping— comes under attack - Data scraping practices for training AI models have come under attack, with OpenAI facing lawsuits alleging unlawful copying of book text and privacy violations, while Twitter has limited access to its data to curb AI data scraping; the debate around data scraping is raising questions about privacy, transparency, and fair use of copyrighted material in AI training.
Google Says It'll Scrape Everything You Post Online for AI - Google has updated its privacy policy to explicitly state that it reserves the right to scrape everything users post online in order to build its AI tools, raising privacy concerns and legal questions about the use of public data.
Anger from voice actors as NSFW mods use AI deepfakes to replicate their voices: 'This is NOT okay.' - Voice actors are expressing anger and frustration as their voices are being used without consent for NSFW mods through AI deepfakes, leading to calls for action and awareness within the voice acting community.
G/O Media’s AI ‘innovation’ is off to a rocky start - G/O Media's early experiments with AI-generated articles have resulted in embarrassing mistakes and criticism from readers and staff.
Sarah Silverman is suing OpenAI and Meta for copyright infringement. - Comedian Sarah Silverman and authors Christopher Golden and Richard Kadrey are suing OpenAI and Meta for copyright infringement, alleging that their works were illegally used to train AI models without their consent.
Man Who Tried to Kill Queen With Crossbow Encouraged By AI Chatbot, Prosecutors Say - A 19-year-old man who attempted to assassinate Queen Elizabeth II with a crossbow was encouraged by an AI chatbot named "Sarai" that engaged in conversations about his plan and expressed support for his actions.
Analysis
AI likely to spell end of traditional school classroom, leading expert says - Advances in AI could lead to the end of traditional school classrooms, with personalized AI tutors potentially delivering high-quality education to every child in the world, according to leading AI expert Prof Stuart Russell, although he cautioned that there are risks involved, including the potential for indoctrination and the need to preserve and improve the social aspects of childhood.
How human translators are coping with competition from powerful AI - AI has not put many human translators out of work due to the specialized nature of certain translation fields and the growth of hybrid translation services, but it has led to lower rates and reduced job opportunities for translators.
Policy
OpenAI is forming a new team to bring ‘superintelligent’ AI under control - OpenAI is forming a new team led by Ilya Sutskever to develop ways to steer and control "superintelligent" AI systems, as they predict that AI with intelligence exceeding that of humans could arrive within the decade.
The Vatican Releases Its Own AI Ethics Handbook - The Vatican has released a handbook on the ethics of artificial intelligence, providing guidance for tech companies on principles such as accountability, transparency, and respect for human dignity and rights.
UN council to hold first meeting on potential threats of artificial intelligence to global peace - The UN Security Council will hold its first-ever meeting on the potential threats of artificial intelligence to international peace and security, organized by the United Kingdom, which sees both the tremendous potential and major risks associated with AI, such as its use in autonomous weapons or control of nuclear weapons.
New York employers to start telling applicants when they encounter AI - Starting this week, job-hunters in New York City will be informed about whether and how artificial intelligence is being used in the hiring process, as part of a new law that requires employers to disclose the use of AI tools and conduct bias audits.
UAE Cabinet approves testing of self-driving vehicles on its roads - The UAE Cabinet has approved the first license for self-driving vehicles to operate on the country's roads, granted to WeRide company, and has also approved a national policy for electric vehicles.
San Francisco may soon get 24/7 driverless cabs. City leaders are fuming. - San Francisco city leaders are unhappy with the idea of driverless taxis operating 24/7 in the city, citing incidents of traffic disruption and safety concerns, but have little say in the matter as it is regulated by the state.
NYC’s New AI Bias Law Broadly Impacts Hiring and Requires Audits - New York City's new AI bias law requires employers to perform annual audits of their recruitment technology to check for instances of bias, with failure to comply resulting in fines, signaling a broader trend of AI bias laws being considered in other states and potentially globally.
‘You can do both’: experts seek ‘good AI’ while attempting to avoid the bad - Experts at the UN's AI for Good global summit discussed the potential benefits and risks of AI, with some calling for a focus on AI that is already in widespread use rather than hypothetical risks, while others emphasized the need to imbue AI with human values and establish governance quickly.
The US Military Is Taking Generative AI Out for a Spin - The US military is exploring the use of generative AI.
Google Updates Privacy Policy To Collect Public Data For AI Training - Google has updated its privacy policy to allow the collection and analysis of public data for training its AI models, raising concerns about privacy and the potential misuse of publicly available information.
ChatGPT pauses Bing integration to stop people from bypassing paywalls - OpenAI has temporarily disabled the Browse with Bing feature in ChatGPT due to users discovering a way to bypass paywalls using the chatbot.
AI Act: Spanish presidency sets out options on key topics of negotiation - The Spanish presidency of the EU Council of Ministers is preparing to negotiate the AI Act, focusing on the topics of AI definition, high-risk classification, list of high-risk use cases, and the fundamental rights impact assessment.
The FTC’s biggest AI enforcement tool? Forcing companies to delete their algorithms - The Federal Trade Commission has been using algorithm disgorgement, or model deletion, as an enforcement tool against tech companies that have used improperly obtained data to build their AI products, requiring them to delete the data and the products developed from it.
Fun
Written in the Stars? More Like Written by A.I. - A machine resembling a standing washer-dryer with celestial buttons and lights allows users to ask questions about their lives, such as how to improve their job or whether to start a cult.
Two guys who were fed up with spam calls created a ChatGPT tool that trolls telemarketers and wastes their time - A telephone service called Jolly Roger uses AI to generate chatty characters that waste the time of telemarketers, keeping them on the line for as long as 30 minutes.
TikTok users are turning AI filters into fortunetellers. - TikTok users are using AI filters as fortunetellers and fate predictors, divining everything from the age of their crush to whether their marriage is meant to last, reflecting a larger cultural desire for artificial intelligence to be more than what it is.
Copyright © 2023 Skynet Today, All rights reserved.