Welcome back to AI Weekly Digest #36, your go-to resource for the most important artificial intelligence news and events from Week 51 of 2023.
As an AI enthusiast, staying informed about the rapid advancements in the field is crucial for both personal and professional growth. Our mission is to bring you a concise and comprehensive roundup of the latest breakthroughs, innovations, and discussions shaping the world of AI.
Dec 23, 2023
Table Of Contents
- AI Conversations: LinkedIn’s Content Analysis Reveals Rising Professional Optimism
- Midjourney V6 is here with in-image text and completely overhauled prompting
- OpenAI in Talks to Raise New Funding at $100 Billion Valuation
- Son Uses AI to Recreate Late Father’s Voice for Special Christmas Present
- Apple Aims to Harness News for AI Model Training
- Pika.art Introduced Expand Canvas Feature
- 23 of Google’s biggest moments in 2023
- New AI Analysis Questions Authenticity of Raphael’s Madonna della Rosa
- Apple Aims to Bring AI Capabilities On-Device Instead of the Cloud
- Google Gemini Falls Short of GPT-3.5 Turbo, New Research Reveals
- Stable Video Diffusion Has Been Added To The Stability AI Developer Platform API
- You can now archive your chats in ChatGPT
- Anthropic will help users if they get sued for copyright infringement
- Scientists discover the first new antibiotics in over 60 years using AI
- AI Robot Physically Outperforms Humans in Astonishing Breakthrough
- AI Image Generators Trained on Child Abuse Images, Study Shows
- Google Unveils VideoPoet: Revolutionizing Zero-Shot Video Generation
- AI Cannot Be Patent ‘Inventor’, UK Supreme Court Rules in Landmark Case
- Brooklyn’s First Chinese-American Councilwoman Elect Leans on AI to Bridge Communication Gap
- GPT-4 driven robot takes selfies, ‘eats’ popcorn
- Microsoft Copilot’s New Tune: Integrating Suno for AI-Powered Music Creation
- Mistral will release an open source Gpt4 level model in 2024
- Google Bard can now support exporting more languages to Replit
- OpenAI’s Strategic Move: Enhancing AI Safety and Empowering Board Oversight
- Research finds people struggle to identify AI from human art, but prefer human-made works
- Could Eye Scans Hold the Key to Early Autism Diagnosis? AI Study Claims 100% Accuracy
- The Rise of AI Fake News: A Misinformation Superspreader
- Gemini Pro is now available for all users on Poe
AI Conversations: LinkedIn’s Content Analysis Reveals Rising Professional Optimism
LinkedIn’s content analysis has unveiled a remarkable shift in professional conversations. As we navigate the ever-evolving landscape of artificial intelligence (AI), this research sheds light on how our interactions are being transformed.
The study, published in Scientific Reports, delves into the impact of AI on communication dynamics. Researchers found that conversations facilitated by AI chat tools exhibit greater efficiency and a surge in positive language. Participants using AI-generated smart replies experienced smoother exchanges, with these automated suggestions accounting for approximately 14.3% of messages sent.
However, there’s a twist. When participants suspected their conversation partner was relying on AI, perceptions changed. The mere suspicion of AI involvement led to negative evaluations. It seems that while AI streamlines our expression, it subtly alters our personal voice, making us sound more positive but potentially sacrificing authenticity.
Midjourney V6 is here with in-image text and completely overhauled prompting
Midjourney, the popular AI art generator, released a major upgrade last night with its version 6 model. Among the improvements are more realistic and detailed images, the ability to generate legible text within images, upgraded image upscalers, and better prompt following. Users will need to re-learn prompting techniques, as the new model is much more sensitive. Developer David Holz advises avoiding previously popular prompts like “photorealistic” and “4K” in favor of more explicit language about desired results.
OpenAI in Talks to Raise New Funding at $100 Billion Valuation
OpenAI, the artificial intelligence powerhouse behind ChatGPT, is currently engaged in early discussions to secure a fresh round of funding. According to sources familiar with the matter, the valuation for this funding round is expected to soar to or exceed $100 billion. While the specifics of the deal are yet to be finalized, the potential implications are monumental.
OpenAI’s negotiation revolves around a transaction known as a tender offer, where existing employees’ shares would be sold at the staggering valuation. The company, which boasts Microsoft Corp. as a 49% owner, is led by Chief Executive Officer Sam Altman and President Greg Brockman. If successful, OpenAI would catapult into the ranks of the world’s most valuable closely held companies, trailing only SpaceX and ByteDance.
Dec 22, 2023
Son Uses AI to Recreate Late Father’s Voice for Special Christmas Present
Phillip Willett recently shared a heartwarming TikTok video that has touched the hearts of many. In this poignant tale, he surprises his mother with a Christmas gift that transcends the ordinary. It’s a gift that bridges the gap between the past and the present, between memory and reality.
The emotional resonance of this gesture is profound. Imagine hearing the familiar timbre, the comforting cadence of a loved one who has departed. Through the magic of artificial intelligence, Phillip has brought his father’s voice back to life, capturing the essence of their shared history.
The technology behind this feat is both remarkable and poignant. It’s a testament to the power of AI, which can now analyze existing voice recordings and create eerily accurate replicas. As we approach the holiday season, this heartwarming story reminds us that love and innovation know no bounds.
Apple Aims to Harness News for AI Model Training
Apple is setting its sights on the news industry to bolster its artificial intelligence (AI) capabilities. According to a recent report, the tech giant plans to leverage news content to enhance its AI models. While Apple has already integrated AI into various aspects of its ecosystem, including photo improvements and text autocorrection, it seeks to catch up with competitors who have made significant strides in AI development.
Pika.art Introduced Expand Canvas Feature
Dec 21, 2023
23 of Google’s biggest moments in 2023
As 2023 comes to a close, Google has made significant strides in AI and technology, marking the year with 23 notable achievements. At the forefront is the introduction of Gemini, Google’s most advanced AI model, boasting multimodal capabilities across various mediums such as text, audio, video, and code. This innovation, available in three versions – Ultra, Pro, and Nano – is a testament to Google’s commitment to making AI accessible on a wide range of devices.
In the realm of interactive AI, Google launched Bard, an experimental platform that has since evolved with coding abilities, visual enhancements, and global expansion. Notably, Bard is poised to integrate with Google Assistant, promising a more seamless user experience. Google also delved into generative AI with its Search Generative Experience (SGE), revolutionizing information discovery and visualization in searches.
New AI Analysis Questions Authenticity of Raphael’s Madonna della Rosa
Art experts have long debated whether Renaissance master Raphael painted the entire Madonna della Rosa, or if his pupil Giulio Romano contributed. Now AI analysis reveals the face of Joseph was likely not painted by Raphael.
Researchers from the Universities of Nottingham, Bradford and Stanford trained an algorithm on authenticated Raphael works. It can identify his style with 98% accuracy. Testing the Madonna della Rosa, it confirmed most as Raphael’s work. However, the face of Joseph was classified as not Raphael’s style.
The study demonstrates AI can be a useful tool in art authentication, alongside traditional methods like provenance research and scientific analysis. It does not replace connoisseurship, but enhances it. Researchers emphasize AI should not make definitive judgments, but contribute evidence for broader art historical interpretation.
Though not conclusive alone, the new analysis strongly suggests Joseph’s face was painted by another hand, possibly Raphael’s pupil Giulio Romano. It sheds intriguing new light on the debated history of a Renaissance masterpiece.
Apple Aims to Bring AI Capabilities On-Device Instead of the Cloud
Apple is actively researching how to run advanced AI models like ChatGPT directly on iPhones and other devices, rather than relying solely on the cloud. A new research paper titled “LLM in a Flash” details an approach to optimize large language models to run efficiently with limited memory, enabling AI-powered features to work offline and with low latency. This signals Apple’s interest in competing with rivals like Google and Samsung in on-device generative AI. Bringing more AI capabilities natively to Apple hardware provides privacy benefits and could help reinvigorate smartphone sales, which declined this year. Technical challenges remain, but the research shows Apple’s progress in overcoming limitations like battery life to make complex AI models practical on personal gadgets. Apple hopes to harness the “full potential” of AI innovations across its products and close the perceived gap with Big Tech competitors in this key technology area.
Google Gemini Falls Short of GPT-3.5 Turbo, New Research Reveals
Researchers from Carnegie Mellon University and BerriAI have conducted a thorough examination, and their findings are clear: Gemini Pro, the most powerful version available to consumers, lags behind OpenAI’s older model, GPT-3.5 Turbo, in most tasks.
The implications are significant. Less than a month after Google’s flashy demo video introduced Gemini, showcasing its capabilities, the reality check is sobering. Despite months of development, Gemini Pro falls short of its rival.
Stable Video Diffusion Has Been Added To The Stability AI Developer Platform API
Dec 20, 2023
You can now archive your chats in ChatGPT
Anthropic will help users if they get sued for copyright infringement
The company updated its commercial terms of service for those who use Claude, its AI chatbot, saying it will not only defend customers from copyright infringement claims but also pay for settlements. The new terms go into effect on January 1st, 2024 and follow similar commitments from Microsoft, Google, and other companies.
These copyright protections won’t apply if a customer “knows or reasonably should know” they’re infringing copyright. This is consistent with what Anthropic previously said in a comment to the US Copyright Office, where it said it didn’t believe “users should be able to create outputs using Claude that infringe copyrighted works.”
Scientists discover the first new antibiotics in over 60 years using AI
In a significant medical breakthrough, researchers at MIT have discovered a new class of antibiotics for treating drug-resistant Staphylococcus aureus (MRSA), achieved through advanced AI deep learning models. This discovery, marking the first in over 60 years, could be pivotal in combating antibiotic resistance. The team used an expanded deep learning model, analyzing around 39,000 compounds for antibiotic activity against MRSA. Their method included assessing compound toxicity on human cells, leading to the identification of two promising antibiotics. This novel approach, detailed in the journal Nature, highlights the intersection of AI technology and pharmaceutical innovation, offering new hope in the fight against drug-resistant infections.
AI Robot Physically Outperforms Humans in Astonishing Breakthrough
Swiss researchers have unveiled an artificial intelligence (AI) robot that not only out-thinks but also physically surpasses human capabilities. The robot, developed by a team in Switzerland, achieved mastery in a popular physical skill game within a mere six hours of learning. This remarkable feat underscores the rapid progress of AI technology, pushing the boundaries of what machines can achieve.
AI Image Generators Trained on Child Abuse Images, Study Shows
A new report reveals that popular AI image generators like Stable Diffusion contain thousands of child sexual abuse images in their training data. Researchers found over 3,200 such illegal images in the LAION dataset, which is used to train leading systems.
The report shows this tainted data makes it easier for AI to generate realistic fake imagery of children and transform social media photos of clothed teens into nudes. In response, LAION temporarily removed its datasets to audit them, though the images represent a fraction of its over 5 billion item index.
Google Unveils VideoPoet: Revolutionizing Zero-Shot Video Generation
Google Research recently introduced VideoPoet, a large language model (LLM) that represents a significant leap in video generation technology. This model addresses a key challenge in the field: producing coherent large motions in videos. Unlike current models that often struggle with larger motions or exhibit artifacts, VideoPoet excels in creating a variety of video outputs with remarkable fluidity and accuracy.
VideoPoet’s abilities include text-to-video, image-to-video, video stylization, video inpainting and outpainting, and even video-to-audio conversion. Its use of multiple tokenizers enables it to encode and decode video, image, and audio clips, allowing for seamless integration of these modalities. This approach avoids the need for separately trained components for each task, setting VideoPoet apart from existing models.
AI Cannot Be Patent ‘Inventor’, UK Supreme Court Rules in Landmark Case
The UK Supreme Court has ruled that artificial intelligence (AI) cannot be recognized as an inventor for patent purposes. The case centered around an American computer scientist, Stephen Thaler, who sought to register patents for inventions generated by his AI system, known as Dabus.
The court’s ruling affirms that patents, which grant ownership rights over new inventions, must originate from human inventors. While Thaler argued that Dabus should be credited as the inventor, the Intellectual Property Office (IPO) rejected his applications. Thaler’s unconventional approach sparked a legal battle that has implications far beyond the courtroom.
At the heart of the matter lies the question of whether existing patent laws, designed for human inventors, can be extended to cover AI-generated innovations. The court’s majority opinion asserted that only a person—rather than a machine—can hold patent rights. Lady Justice Elisabeth Laing emphasized that “a patent is a statutory right and can only be granted to a person.”
Dec 19, 2023
Brooklyn’s First Chinese-American Councilwoman Elect Leans on AI to Bridge Communication Gap
Susan Zhuang, soon to be Brooklyn’s first Chinese-American Councilwoman, has sparked both praise and criticism after admitting to using AI to answer questions from a local news outlet. Zhuang, a Democrat representing the 43rd District, defended her decision, stating that as an immigrant with English as a non-native language, AI serves as a valuable tool for fostering deeper understanding and personal growth.
GPT-4 driven robot takes selfies, ‘eats’ popcorn
In a delightful fusion of artificial intelligence and whimsy, researchers at OpenAI have unveiled their latest creation: a GPT-4 driven robot that’s not only adept at snapping selfies but also indulges in a peculiar pastime—”eating” popcorn.
Microsoft Copilot’s New Tune: Integrating Suno for AI-Powered Music Creation
Microsoft Copilot, an AI chatbot, now boasts a music creation feature through its integration with the GenAI music app Suno. This allows users to generate complete songs, including lyrics and instrumentals, from simple prompts. The feature is accessible via Microsoft Edge and the Copilot website, marking a significant advancement in AI-driven music technology. However, the ethical and legal concerns surrounding AI-synthesized music, particularly around copyright and artists’ rights, remain unresolved. This development highlights both the creative potential and the complexities of AI in the music industry.
Mistral will release an open source Gpt4 level model in 2024
Dec 18, 2023
Google Bard can now support exporting more languages to Replit
OpenAI’s Strategic Move: Enhancing AI Safety and Empowering Board Oversight
OpenAI is taking significant steps to bolster its AI safety measures. The establishment of a new Safety Advisory Group, designed to provide guidance to the leadership team, marks a pivotal move in the organization’s approach to managing the risks associated with artificial intelligence. This group is set to play a critical role in overseeing the development of AI models, particularly in assessing and addressing catastrophic risks that could potentially lead to extensive economic damage or harm to individuals.
In response to recent organizational changes, including the departure of key board members, OpenAI has updated its “Preparedness Framework”. This framework is vital for identifying, analyzing, and making decisions about the severe risks inherent in the AI models they develop. OpenAI categorizes these risks into four primary areas: cybersecurity, persuasion (such as misinformation), model autonomy, and CBRN (chemical, biological, radiological, and nuclear threats). The assessment of these risks is crucial in determining whether a model is safe to deploy or requires further development.
An intriguing aspect of this initiative is the delegation of veto power to OpenAI’s board. This move allows the board to overturn decisions regarding the deployment of high-risk AI products and processes. This change aims to prevent scenarios where high-risk AI developments proceed without board consent, enhancing the organization’s transparency and accountability. However, questions remain about the effectiveness of this strategy, particularly regarding the board’s readiness to challenge decisions and the level of transparency in these processes. OpenAI’s commitment to soliciting independent third-party audits is a step towards addressing these concerns.
Research finds people struggle to identify AI from human art, but prefer human-made works
A new study reveals that humans struggle to distinguish art created by artificial intelligence (AI) from art made by human hands. But there’s a twist: even when stumped, folks still favor the flesh-and-blood brushstrokes.
Researchers examined how people reacted to both AI-generated and human-made art, finding little difference in their ability to spot the source. This suggests AI artistry has reached a level of sophistication that mimics the human touch, blurring the lines between creative minds and machine algorithms.
But here’s the kicker: despite the close call in the identification game, human-crafted art emerged as the clear winner in the hearts (and minds) of the study participants. This preference, the researchers speculate, might stem from the inherent value we place on the human experience and its expression through art. Whether it’s the emotional brushstrokes or the invisible ink of personal history embedded in each creation, there’s something uniquely captivating about art born from the human soul.
Could Eye Scans Hold the Key to Early Autism Diagnosis? AI Study Claims 100% Accuracy
A groundbreaking study suggests a radical new approach to diagnosing autism in children: analyzing retinal photographs with a deep learning algorithm. Researchers report a startling 100% accuracy rate in identifying autism spectrum disorder (ASD) using this method, potentially paving the way for earlier intervention and improved outcomes.
The study, conducted by scientists at the Icahn School of Medicine at Mount Sinai, involved analyzing retinal images from over 200 children, half with ASD and half without. The AI algorithm, trained on these images, was able to pinpoint subtle differences in retinal structure and blood vessel patterns, accurately distinguishing autistic children from their neurotypical counterparts.
Dec 17, 2023
The Rise of AI Fake News: A Misinformation Superspreader
In a digital landscape increasingly shaped by artificial intelligence (AI), a concerning trend is emerging: the proliferation of AI-generated fake news. The very technology designed to enhance our lives is now being exploited to disseminate false information about wars, elections, and critical global events.
Artificial intelligence is automating the creation of deceptive content, leading to an explosion of websites that peddle misinformation. These platforms, often operated by malicious actors, exploit AI’s capabilities to produce realistic yet entirely fabricated news stories. The consequences are far-reaching, affecting not only public opinion but also the very fabric of our democratic processes.
Gemini Pro is now available for all users on Poe
Every week, we’ll meticulously curate a selection of stories from top AI media outlets and distill them into a digestible format, ensuring you stay up-to-date with the latest developments without having to spend hours browsing the web. From groundbreaking research to real-world applications, ethical debates to policy implications, AI Weekly Digest will be your essential guide to navigating the ever-evolving landscape of artificial intelligence. Join us on this exciting journey as we explore the future of AI together, one week at a time.