Last Week in AI #233: Gen AI at peak of inflated expectations, NYT bans AI vendors from scraping its data, FEC may limit AI political ads before 2024, Hollywood boosts Gen AI spend amid strikes!

Gartner rates Gen AI at peak of inflated expectations, NYT prevents scraping its data for AI, FEC may limit AI political ads before 2024, Hollywood studios boosts Gen AI spend amid strikes

Top News

Gartner Places Generative AI on the Peak of Inflated Expectations on the 2023 Hype Cycle for Emerging Technologies

Gartner's 2023 Hype Cycle for Emerging Technologies places generative AI at the peak of inflated expectations, highlighting its potential for enhancing digital customer experiences and business decisions. The report identifies four key emerging technology trends: Emergent AI, Developer Experience (DevX), Pervasive Cloud, and Human-centric Security and Privacy. Emergent AI includes techniques like AI simulation, causal AI, and federated machine learning. DevX focuses on improving interactions between developers and their tools, with technologies like AI-augmented software engineering and API-centric SaaS. The Pervasive Cloud trend predicts cloud computing becoming an essential driver of business innovation, enabled by technologies like cloud development environments and cloud sustainability. Human-centric Security and Privacy emphasizes the need for a security and privacy program that integrates into an organization's digital design, supported by technologies like AI TRISM and postquantum cryptography.

The New York Times prohibits AI vendors from devouring its content

The New York Times (NYT) has updated its terms of service to prohibit the scraping of its content for AI training, a practice used by tech companies to enhance the capabilities of AI language apps like ChatGPT and Google Bard. The new terms explicitly forbid the use of NYT content, including articles, videos, images, and metadata, for training any AI model without express written permission. The move comes amid controversy over the use of scraped data for AI training, which has led to legal challenges, including a lawsuit accusing OpenAI of plagiarism. OpenAI has begun taking steps to address these concerns, detailing a method for websites to block its AI-training web crawler. However, it remains to be seen whether AI vendors will respect these new terms, with potential legal and regulatory consequences looming.

FEC could limit AI in political ads ahead of 2024

The Federal Election Commission (FEC) is considering implementing rules to regulate the use of AI-generated content in political ads. The decision follows a petition from the advocacy group Public Citizen, which calls for the FEC to create rules that prevent candidates and political parties from using AI to misrepresent their opponents. The FEC has opened the petition for public comment, a process that could result in new rules being in place by the end of the year. However, there are concerns about whether the FEC has the necessary authority to implement these rules, with some suggesting that federal lawmakers should take action first. This comes amid the increasing use of AI in political campaigns and a growing push from Congress and the White House to regulate the technology.

As Hollywood Strikes, 96% Of Entertainment Companies Are Boosting Generative AI Spend

As the Hollywood writer's strike surpasses 100 days, nearly all entertainment and media companies plan to increase expenditures on generative AI. This stems from a report by Lucidworks indicating that 96% of global executives involved in AI decisions prioritize generative AI investments. The trend is leading globally, with 100% of companies in China, 98% in India, 95% in France, 94% in the UK, and 92% in the USA investing in generative AI. Amid fears of AI's potential impact on professions, the entertainment industry is rapidly moving to incorporate or purchase generative AI technology. This could lead to 90% of Hollywood content being at least partly AI-generated by 2025. Furthermore, the emergence of synthetic actors through AI technologies could pose a genuine threat to the industry's talent pool.

Other News

Research

DeepMind’s AlphaStar Benchmark Improves RL Offline Agent With 90% Win Rate Against SOTA AlphaStar Supervised Agent - DeepMind's AlphaStar Unplugged presents a challenging offline reinforcement learning benchmark using StarCraft II, with their baseline offline agent achieving a 90% win rate against the previous state-of-the-art AlphaStar supervised agent.

How machine learning models can amplify inequities in medical diagnosis and treatment - Researchers at MIT have investigated the fairness of machine learning models under subpopulation shift, identifying four types of shifts that can lead to biases and disparities in medical diagnosis and treatment, and highlighting the need for more equitable models.

Deep learning-based AI model can estimate age from chest radiographs of healthy individuals - Osaka Metropolitan University scientists have developed an advanced AI model that accurately estimates a patient's chronological age using chest radiographs, which can also signal a correlation with chronic diseases, paving the way for improved early disease detection and intervention.

MIT researchers combine deep learning and physics to fix motion-corrupted MRI scans - MIT researchers have developed a deep learning model capable of motion correction in brain MRI scans, which could help reduce image artifacts and improve diagnoses.

Follow Anything: Open-set detection, tracking, and following in real-time - A robotic system called "follow anything" (FAn) has been developed to detect, track, and follow any object in real-time using multimodal queries, and it can be deployed on a laptop with a lightweight graphics card.

RAVEN: In-Context Learning with Retrieval Augmented Encoder-Decoder Language Models - RAVEN is an in-context learning model that utilizes retrieval augmented encoder-decoder language models for improved performance.

Automatically Correcting Large Language Models: Surveying the landscape of diverse self-correction strategies - A survey of diverse self-correction strategies for large language models in the context of AI.

Announcing AI2 OLMo, an open language model made by scientists, for scientists - The Allen Institute for AI is creating an open language model called AI2 OLMo, which will be accessible to the research community and will provide access to all aspects of model creation, including data, code, and ethical considerations.

Risky Giant Steps Can Solve Optimization Problems Faster - Advances in computer-aided proof techniques have led to the discovery that taking big steps in the gradient descent algorithm can solve optimization problems faster than the conventional wisdom of taking small steps.

Sketches in Computer Vision - Sketches are gaining attention in the field of computer vision as a unique and illustrative mode of visually representing natural entities, and this article provides an overview of sketch-based computer vision, including sketch representation learning and its application in various computer vision tasks.

Can machine learning algorithms identify patients at risk of a delay in starting cancer treatment? - Machine learning models incorporating clinical and demographic data of cancer patients and neighborhood-level social determinants of health data can effectively identify patients at risk of experiencing treatment delays of more than 60 days after their initial cancer diagnosis, although the model exhibits lower predictive effectiveness in vulnerable populations.

LLaMA-2 from the Ground Up - LLaMA-2 is an open-source large language model that outperforms other models in terms of size, pre-training data, and architecture, and it focuses on alignment and safety during the fine-tuning process to create LLaMA-2-Chat models optimized for dialogue applications.

CoDeF: Content Deformation Fields for Temporally Consistent Video Processing - CoDeF is a new video representation that uses content deformation fields to optimize video reconstruction and support image algorithms for video processing, achieving superior cross-frame consistency and the ability to track non-rigid objects without training.

ChatGPT Isn't As Good At Coding As We Thought - ChatGPT's coding skills are underwhelming, with 52% of its answers containing inaccuracies and 77% being verbose, according to a study from Purdue University.

Self-Alignment with Instruction Backtranslation - The article discusses the process of self-alignment with instruction backtranslation in the context of AI.

Applications

Google Tests an A.I. Assistant That Offers Life Advice - Google is testing an AI assistant that can answer intimate questions about people's lives and offer suggestions and recommendations based on different situations, although the company's AI safety experts have warned about the potential negative effects of relying too much on AI for life advice.

Google Photos adds a scrapbook-like Memories view feature aided by AI - Google Photos introduces a new Memories view feature that allows users to create a scrapbook-like timeline of their favorite moments, with the help of AI-generated titles and the ability to collaborate with others, and share the timeline on social media platforms.

Google Chrome will summarize entire articles for you with built-in generative AI - Google Chrome's new feature, SGE while browsing, uses generative AI to summarize articles and provide key points for users, making it easier to find relevant information.

How To Swap Your Face Into Midjourney Images (InsightFaceSwap / Picsi.AI Demo) - A tool called Picsi.AI allows users to swap their faces into Midjourney images on Discord, with the free tier offering 50 swap credits a day.

Business

OpenAI acquires AI design studio Global Illumination - OpenAI has acquired Global Illumination, a New York-based startup that uses AI to build creative tools and digital experiences, marking OpenAI's first public acquisition in its history.

AI startup Anthropic raises $100M from Korean telco giant SK Telecom - Anthropic, an AI startup co-founded by former OpenAI leaders, has secured $100 million in funding from South Korean mobile carrier SK Telecom (SKT) to co-develop a multilingual large language model customized for global telco firms.

Microsoft Azure ChatGPT allows enterprises to run ChatGPT within their network - Microsoft Azure ChatGPT allows enterprises to run ChatGPT within their network, providing a private and controlled AI solution with added benefits such as privacy guarantees and integration with internal services.

The scientist behind IBM Watson has raised $60 million for his AI startup in New York - David Ferrucci, the scientist behind IBM Watson, has raised $60 million for his AI startup Elemental Cognition, which aims to develop AI that "thinks before it talks" and offers chatbot products for various industries.

What you need to know about Sakana AI, the new startup from a transformer paper co-author - Sakana AI, a new AI company founded by Llion Jones and David Ha, is pursuing a biomimicry approach to AI by designing models inspired by collective intelligence in nature, with the goal of creating a generative AI model proficient in generating text, images, code, and multimedia content.

Dialpad launches generative AI trained on 5 years of proprietary conversational data - Dialpad has launched DialpadGPT, a generative AI model specifically designed for customer service, sales, and recruiting, which automates tasks such as summarizing sales, customer service, and team activities.

Voiceflow, a platform for building conversational AI experiences, raises $15M - Voiceflow, a platform for building conversational AI experiences, has raised $15 million in funding to enhance its product innovation, including the addition of an AI builder for customer support and more.

Robomart is banking on ‘store-hailing’ to bring self-driving stores directly to customers - Robomart, a self-driving store company, has raised $2 million in funding and unveiled a new product called Haven, which offers retailers customized branding options and the capacity for around 300 SKUs, allowing supermarkets to expand their retail footprint and provide consumers with a full in-person shopping experience at home.

Adobe Express now has AI-powered features to take on Canva - Adobe Express, a cloud-based design platform, is rolling out AI-powered features including custom text and image effects, automatic background removal, and character animations, making content creation more accessible for users without professional design experience.

Have you used generative AI to shop? 17% of shoppers have - Generative AI is rapidly growing in the retail industry, with 17% of shoppers using AI for purchasing inspiration and 45% interested in using it as a shopping resource, according to a new report by Salesforce.

40% of workers will have to reskill in the next three years due to AI, says IBM study - 40% of workers will need to reskill in the next three years due to AI implementation, but a new IBM study shows that rather than fearing the technology, people should leverage it for their own gain, as executives expect generative AI to augment roles rather than replace them.

NCSoft's new AI suite is trained to streamline game production - NCSoft has developed four new AI large language models, called VARCO, to streamline game development, with potential products including digital humans, generative AI platforms, and conversational language models.

Former Google DeepMind Employees Venture into Startups, AI, and More - Over 200 former employees of Google DeepMind have left the organization to start their own startups in various fields, such as AI, cryptocurrency, biotechnology, and climate tech, with the support and guidance of DeepMind's executive team who are active angel investors in AI startups.

Why Google’s co-founder left retirement to work on AI - Google's co-founder returned from retirement to work on AI, as he made a surprise appearance at an AI event and explained his interest in the field.

Ex-MZ CEO launches BeFake, an AI-based social media app - Alias Technologies, an AI company started by former Machine Zone CEO Kristen Garcia Dumont, has launched BeFake, a social media app that allows users to transform themselves into captivating, augmented visuals and explore the endless possibilities of their creativity beyond the limitations of the physical world.

Runway launches new ‘Watch’ feature as CEO says Hollywood AI discourse ‘needs to be more nuanced’  - Runway, a gen AI video startup, has launched a new feature called Watch, which allows users to share and consume longer-form videos created with Runway tools, as CEO Cristóbal Valenzuela emphasizes the need for a more nuanced conversation around Hollywood and AI.

Thomson Reuters Completes Acquisition of Casetext, Inc. - Thomson Reuters has completed its acquisition of Casetext, a provider of AI technology for legal professionals, for $650 million in cash, as part of its efforts to bring generative AI solutions to its customers and redefine the future of professionals through applications of generative AI.

New $15 Million Fund Launched for AI Startups - Tech veteran Chris Messina and writer/podcaster Brian McCullough have launched a $15 million fund to support AI startups, attracting investments from prominent tech luminaries, with a focus on generative AI and niche industries.

Concerns

These Women Tried to Warn Us About AI - Researchers have been warning about the biases and dangers of AI systems for years, particularly in relation to issues of race and gender, but their concerns have often been ignored or dismissed by the industry; however, as AI becomes more prevalent and its impact on society more apparent, governments and industry leaders are finally starting to take notice and address the need for regulation and accountability.

Yoshua Bengio: Democracy is not safe in an AI world - Yoshua Bengio warns about the necessity of regulations on AI technology, discussing its potential to be used for harm and its impact on humanity.

AI isn’t great at decoding human emotions. So why are regulators targeting the tech? - Regulators are targeting AI technology that decodes human emotions due to concerns about its potential misuse for surveillance and social control.

When Hackers Descended to Test A.I., They Found Flaws Aplenty - Hackers at the Defcon conference tested the vulnerabilities of various AI programs, attempting to break through their safeguards and identify flaws before criminals and misinformation peddlers could exploit them.

AI Detection Tools Falsely Accuse International Students of Cheating – The Markup - AI detection tools used by academic institutions to spot plagiarized text and flag AI-generated writing are falsely accusing international students of cheating due to a bias against non-native English speakers, as their writing tends to be simpler and more predictable, which is often flagged as AI-generated.

A.I. use rising in influence campaigns online, but impact limited - US cyber firm - The use of AI in manipulative information campaigns online has been increasing, but its impact has been limited so far, according to a report by cybersecurity firm Mandiant.

A.I. use rising in influence campaigns online, but impact limited - US cyber firm - The use of AI in manipulative information campaigns online has been increasing, but its impact has been limited so far, according to a report by cybersecurity firm Mandiant.

A.I. founder charged with defrauding investors, manipulating documents - The founder of a data analytics firm that claimed to use AI technology has been indicted for defrauding investors by manipulating financial documents to create a false impression of success.

AI chatbot scares Snapchat users by posting mysterious video - Snapchat's AI chatbot, My AI, posted a mysterious one-second video of a wall and ceiling, causing users to speculate about its intentions and whether it was spying on them.

Driverless Car Gets Stuck in Wet Concrete in San Francisco - A driverless car in San Francisco got stuck in wet concrete, raising doubts about the promised benefits of autonomous vehicles.

Tesla knew Autopilot's limitations killed a driver – but didn't fix it, engineers claim - Tesla's Autopilot engineers claim that the company's leadership knew about the software's inability to detect and respond to cross traffic, but did nothing to fix it, according to allegations made in a civil lawsuit brought against Tesla regarding a fatal crash.

Analysis

Does ChatGPT have a liberal bias? - The article discusses a paper that claims ChatGPT, an AI language model, expresses liberal opinions, but the author argues that the paper provides little evidence to support this claim and highlights flaws in the research methodology.

Meta, OpenAI, Anthropic and Cohere A.I. models all make stuff up — here's which is worst - Microsoft-backed OpenAI's GPT-4 performs the best in terms of hallucinations and math accuracy among top AI models, while Meta's Llama 2 hallucinates the most and Anthropic's Claude 2 excels in self-awareness, according to a report from Arthur AI.

What Americans Know About AI, Cybersecurity and Big Tech - Most U.S. adults have a basic understanding of digital topics such as secure passwords and cookies, but fewer are knowledgeable about two-factor authentication and artificial intelligence, with educational attainment and age playing a role in digital literacy.

Policy

The Associated Press sets AI guidelines for journalists - The Associated Press has released guidelines for its journalists on how to use generative AI, cautioning them to treat AI-generated content as unvetted source material and not to use it for publishable content, while also signing an agreement with ChatGPT maker OpenAI to use its news stories to train generative AI models.

OpenAI proposes a new way to use GPT-4 for content moderation - OpenAI proposes using GPT-4 for content moderation by prompting the model with a policy and refining it through feedback from policy experts, claiming that it can reduce the time it takes to roll out new content moderation policies down to hours, although the company acknowledges the need for careful monitoring and validation due to potential biases and mistakes made by AI models.

What normal Americans — not AI companies — want for AI - A majority of American voters want to slow down the development of AI and do not trust AI companies to self-regulate, according to a new poll, indicating a divergence between elite and popular opinion on the technology.

An Iowa school district is using AI to ban books - An Iowa school district is using AI to flag and potentially ban books that do not adhere to new state education curriculum rules.

Expert Opinions

Charlie Kaufman: If Hollywood Wants ‘Garbage,’ They ‘Might as Well Have A.I.’ Write Scripts - Charlie Kaufman criticizes the rise of "garbage" content in Hollywood and warns that if studios use AI to write screenplays, there will be no hope for quality films because AI cannot create moments of humanity.

Fun

In the Battle Between Bots and Comedians, A.I. Is Killing - Artificial intelligence is making strides in the field of comedy, but the closer it gets to imitating human humor, the creepier it becomes, leaving human comedians with an advantage.

---

Copyright © 2023 Skynet Today, All rights reserved.