Last Week in AI #221: Microsoft brings AI to Windows, Adobe adds generative AI to Photoshop, Meta introduces speech-to-text and text-to-speech for 1k+ languages, and more!

More product announcements by big tech highlight incumbent advantage in deploying AI, Meta uses the Bible to train real-time translation AI across 1000+ languages

Top News

Bringing the power of AI to Windows 11 – unlocking a new era of productivity for customers and developers with Windows Copilot and Dev Home

Microsoft is introducing Windows Copilot, a centralized AI assistant for Windows 11, making it the first PC platform to offer such a feature. Windows Copilot will help users take action, customize settings, and connect across their favorite apps. The company is also extending Bing Chat plugins to Windows, enabling developers to integrate their apps within Windows Copilot. Additionally, Microsoft is introducing a new Hybrid AI loop to support AI development across platforms and Azure, with new silicon support from AMD, Intel, Nvidia, and Qualcomm.

Our Take: The introduction of Windows Copilot marks a significant step forward in integrating AI assistance into everyday computing tasks. By centralizing AI capabilities within the operating system, Microsoft is making it easier for users to access and benefit from AI-powered features. This move also opens up new opportunities for developers to create innovative applications that leverage AI assistance, potentially leading to a new wave of productivity tools and experiences.

Adobe is adding AI image generator Firefly to Photoshop

Adobe Photoshop is introducing a new generative AI tool called Generative Fill, which uses Adobe's AI image generator Firefly to extend images and add or remove objects using text prompts. The feature offers more control to users compared to the existing Content-Aware Fill feature. Generative Fill is currently in beta and will see a full release in Photoshop later this year. Adobe claims that the model is trained on content it has the right to use, making the generated images safe for commercial use.

Our Take: The integration of AI image generation tools like Firefly into mainstream applications like Photoshop is a step forward in making these advanced technologies accessible to a wider audience. The emphasis on using legally safe training data is a crucial aspect that addresses concerns about potential legal repercussions when using AI-generated content. Both this story and the previous one about Microsoft highlight yet again the importance of integration of AI into existing products and the advantages of incumbents, who already have a large userbase and platform to surface new AI capabilities.

Introducing speech-to-text, text-to-speech, and more for 1,100+ languages

Meta has introduced the Massively Multilingual Speech (MMS) project, which combines wav2vec 2.0 and a new dataset to provide speech-to-text, text-to-speech, and language identification for over 1,100 languages. The project aims to make information more accessible and help preserve language diversity. The MMS models outperform existing models and cover ten times as many languages. The dataset was created using readings of the New Testament in over 1,100 languages, providing an average of 32 hours of data per language. The models were trained on 500,000 hours of speech data in over 1,400 languages, and the results show good performance compared to current speech models.

Our take: By supporting over 1,100 languages, MMS has the potential to help preserve endangered languages and promote cultural diversity. However, it's essential to consider the limitations of the dataset, which is primarily based on religious texts and may not be representative of everyday language use. Additionally, the focus on male speakers in the dataset could lead to potential biases in the models. Despite these limitations, the MMS project is a promising development in the field of speech recognition and synthesis, and there are many potential applications in VR/AR.

---

Last Week in AI is Seeking Sponsors

Join us in our mission to demystify and disseminate AI news without hype! By sponsoring our newsletter, you're reaching a vibrant, high-value audience of AI enthusiasts, executives, and engineers. We offer tasteful and minimalist ads positioned at the top of each email, ensuring maximum exposure and valuable engagement. Shoot us an email if you’re interested!

---

Other News

Research

Voyager: An Open-Ended Embodied Agent with Large Language Models - “I’m excited to announce Voyager, the first lifelong learning agent that plays Minecraft purely in-context. Voyager continuously improves itself by writing, refining, committing, and retrieving *code* from a skill library.”

Whats the next word in large language models? - "We are trying to keep up with the torrent of developments and discussions in AI and language models since ChatGPT was unleashed on the world."

InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning - "General-purpose language models that can solve various language-domain tasks have emerged driven by the pre-training and instruction-tuning pipeline."

Researchers use AI to identify similar materials in images - "A robot manipulating objects while, say, working in a kitchen, will benefit from understanding which items are composed of the same materials."

Some Neural Networks Learn Language Like Humans - "Researchers uncover striking parallels in the ways that humans and machine learning models acquire language skills. "

LIMA: Less Is More for Alignment - "Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them."

CoDi: Any-to-Any Generation via Composable Diffusion - "We present Composable Diffusion (CoDi), a novel generative model capable of generating any combination of output modalities, such as language, image, video, or audio, from any combination of input modalities."

PandaGPT: One Model to Instruction-Follow Them All - "PandaGPT is a general-purpose instruction-following model that can both see and hear. Our pilot experiments show that PandaGPT can perform complex tasks such as detailed image description generation, writing stories inspired by videos, and answering questions about audios."

BLOOMChat: a New Open Multilingual Chat LLM - "Create a survey that asks about eating and cooking habits. Overall, this type of survey can give you valuable insights into your target audience’s eating and cooking habits which can inform decisions around product development, marketing strategies, recipe development, etc."

Control-A-Video: Controllable Text-to-Video Generation with Diffusion Models - "This paper presents a controllable text-to-video (T2V) diffusion model, named Video-ControlNet, that generates videos conditioned on a sequence of control signals, such as edge or depth maps."

Applications

TikTok is testing an AI chatbot called Tako - "TikTok is testing an AI chatbot called Tako that can recommend videos based on what people ask it, according to screenshots of the feature shared with The Verge."

Google’s open-source AI tool let me play my favorite Dreamcast game with my face - "While Wednesday’s Google I/O event largely hyped the company’s biggest AI initiatives, the company also announced updates to the machine learning suite that powers Google Lens and Google Meet features like object tracking and recognition, gesture control, and of course, facial detection."

A.I. May Help Design Your Favorite Video Game Character - "Intrigued by the potential that generative artificial intelligence holds for video game design, the studio Blizzard Entertainment has trained an image generator on its own hit titles."

How this S’pore AI startup is transforming the marketing sector by automating content creation - "ChatGPT set the tech world ablaze last November and since then, the generative AI train has shown no signs of slowing down. From writing verses to creating album art, AI tools are proving their use across all areas of the creative process."

Business

Google introduces Product Studio, a tool that lets merchants create product imagery using generative AI - "At its Google Marketing Live event today, the tech giant announced that it’s launching Product Studio, a new tool that lets merchants easily create product imagery using generative AI."

Amazon is focusing on using A.I. to get stuff delivered to you faster - "Amazon is focusing on using artificial intelligence to speed up deliveries — by minimizing the distance between its products and customers, a top executive told CNBC."

Broad, Open HPC+AI Portfolio Powers Performance, Generative AI for Science - "At ISC’23, Intel details competitive performance for diverse HPC and AI workloads, from memory-bound to generative AI, and introduces new science LLM initiative to democratize AI."

Hedge funds piled into these A.I. stocks during the first quarter - "Major investors including David Tepper and Bill Ackman scooped up companies capitalizing on the artificial intelligence arms race during the first quarter."

Bing is now the default search for ChatGPT - "OpenAI will start using Bing as the default search experience for ChatGPT. The new search functionality will be rolling out to ChatGPT Plus users today and will be enabled for all free ChatGPT users soon through a plug-in in ChatGPT."

Sequoia India’s Surge backs AI-powered video creation platform Gan.ai in $5.2M funding - "As many as 91% of consumers want to see more videos from brands, according to a recent survey by the startup Wyzowl, which says it surveyed more than 500 respondents. Google meanwhile reported in 2019 that 55% of consumers use videos in their purchasing decisions."

Self-driving car startup Haomo.ai unveils low price delivery robot - "On Monday, Great Wall Motor-backed Chinese self-driving car startup Haomo.ai unveiled the latest generation of its autonomous delivery vehicle, priced at RMB 89,999 ($12,767), which the company claims makes it the cheapest in the industry."

Apple looks to advance artificial intelligence for iPhones, iPads, and other devices - "Apple has posted dozens of job listings in recent months focused on artificial intelligence (AI), signaling that the company may be stepping up its efforts in a field that could transform its signature products."

Nvidia and Microsoft team up to bring generative AI to Windows - "Today at the Microsoft Build developer conference, NVIDIA and Microsoft showcased a suite of advancements in Windows 11 PCs and workstations with NVIDIA RTX GPUs they say will meet the demands of generative AI."

Uber to offer autonomous ride-hailing and delivery using Waymo's robotaxis - "Uber's ride-hailing and delivery services will soon be offered by Waymo's autonomous vehicles -- at least in the Phoenix, Arizona metropolitan area."

Moon Surgical Snags $55.5M for Its Robotic Laparoscopy Assistant - "Robotics startup Moon Surgical recently gained a healthy influx of capital and a distinguished board member. The company, based in Paris and San Francisco, announced the close of a $55.5 million funding round on Wednesday."

That podcast ad you're listening to may soon be AI. Spotify is reportedly developing bots to mimic your favorite hosts. - "Next time you're playing a podcast on Spotify, listen really close to its ad — it may be read by AI, not your favorite host."

UK funds startup Lumai for optical AI - "UK startup company Lumai Ltd. (Oxford, England) has been awarded a £1.1 million grant by the UK government to commercialize its work on all-optical neural networks. "

Stability AI open sources its AI-powered design studio - "Stability AI, the AI startup behind the text-to-image model Stable Diffusion, this week announced the release of StableStudio, an open source version of DreamStudio, Stability AI’s commercial AI-powered design suite."

The A.I. chip boom is pushing Nvidia toward $1 trillion, but it won't help Intel and AMD - "Sales are up because of spiking demand for the graphics processors (GPUs) that Nvidia makes, which power AI applications like those at Google, Microsoft, and OpenAI."

Apptronik readies its humanoid robot for a summer unveil - "Jeff Cardenas pulls out his MacBook. Apptronik’s co-founder and CEO has a slideshow he wants to show, running down the Austin startup’s seven-year history. It does, indeed, take a bit of contextualizing."

JPMorgan is developing a ChatGPT-like A.I. service that gives investment advice - "JPMorgan Chase is developing a ChatGPT-like software service that leans on a disruptive form of artificial intelligence to select investments for customers, CNBC has learned. The company applied to trademark a product called IndexGPT this month, according to a filing from the New York-based bank."

Universal Music Group Signs Deal With AI Startup Endel - "Universal Music Group (UMG) has announced a partnership with Endel, a company that uses AI to generate soundscapes meant to facilitate work, meditation, and sleep."

AI startup Figure raises $70 million to build humanoid robots - "Figure, a startup that wants to build general-purpose humanoid robots, has raised $70 million from investors led by Parkway Venture Capital in its first external round, the company said on Wednesday."

Concerns

Big tech says it can boost productivity, but AI won’t solve meetings madness - "Endless Zoom meetings, a full inbox, innumerable Slack notifications – who’s got any time to actually work?"

Google’s new Magic Editor pushes us toward AI-perfected fakery - "Google’s AI-powered Magic Editor tool promises to ‘make complex edits without pro-level editing tools.’"

A Milton resident’s lawsuit against CVS raises questions about the use of AI lie detectors in hiring - "It’s illegal for employers in Massachusetts to use a lie detector to screen job applicants, but what if a company uses artificial intelligence to help assess a candidate’s honesty?"

Deepfake' scam in China fans worries over AI-driven fraud - "A fraud in northern China that used sophisticated "deepfake" technology to convince a man to transfer money to a supposed friend has sparked concern about the potential of artificial intelligence (AI) techniques to aid financial crimes."

AI scanner used in hundreds of US schools misses knives - "Image source, EvolvBBC North America technology reporterA security firm that sells AI weapons scanners to schools is facing fresh questions about its technology after a student was attacked with a knife that the $3.7m system failed to detect."

Fake AI-generated image of explosion near Pentagon spreads on social media - "An AI-generated image that appeared to show an explosion next to a building in the Pentagon complex circulated on social media platforms on Monday, in the latest incident to highlight concerns over misinformation generated by AI."

An A.I.-Generated Spoof Rattles the Markets - "For a few minutes on Monday, an ominous image of black smoke billowing from what appeared to be a government building near the Pentagon set off investor fears, sending stocks tumbling."

A.I. poses existential risk of people being 'harmed or killed,' ex-Google CEO Eric Schmidt says - "Artificial intelligence could pose existential risks and governments need to know how to make sure the technology is not "misused by evil people," former Google CEO Eric Schmidt warned Wednesday."

An early warning system for novel AI risks - "To pioneer responsibly at the cutting edge of artificial intelligence (AI) research, we must identify new capabilities and novel risks in our AI systems as early as possible."

Analysis

OpenAI vs. Anthropic is the Key 2023 Startup Battle to Watch in Generative A.I. - "OpenAI has gone on a PR splurge about A.I. regulation, but it’s not ASI (artificial super intelligence) that is its biggest competitor, it’s Anthropic. "

Policy

Biden-⁠Harris Administration Takes New Steps to Advance Responsible Artificial Intelligence Research, Development, and Deployment - "Today, the Biden-Harris Administration is announcing new efforts that will advance the research, development, and deployment of responsible artificial intelligence (AI) that protects individuals’ rights and safety and delivers results for the American people."

OpenAI is exploring collective decisions on AI, like Wikipedia entries - "AI is developing rapidly enough and the dangers it may pose are clear enough that OpenAI’s leadership believes that the world needs an international regulatory body akin to that governing nuclear power — and fast. But not too fast."

OpenAI leaders propose international regulatory body for AI - "AI is developing rapidly enough and the dangers it may pose are clear enough that OpenAI’s leadership believes that the world needs an international regulatory body akin to that governing nuclear power — and fast. But not too fast."

Israel aims to be 'AI superpower', advance autonomous warfare - " Israel aims to parlay its technological prowess to become an artificial intelligence "superpower", the Defence Ministry director-general said on Monday, predicting advances in autonomous warfare and streamlined combat decision-making."

Apple reportedly limits internal use of AI-powered tools like ChatGPT and GitHub Copilot - "As big tech companies are in a fierce race with each other to build generative AI tools, they are being cautious about giving their secrets away."

EU, Google to develop voluntary AI pact ahead of new AI rules, EU's Breton says - "Alphabet (GOOGL.O) and the European Commission aim to develop an artificial intelligence (AI) pact involving European and non-European companies ahead of rules to govern the technology, EU industry chief Thierry Breton said on Wednesday."

ChatGPT-maker warns it might leave EU over planned AI law - "The boss of the company behind ChatGPT has said it might consider leaving the EU if it fails to comply with a planned law on artificial intelligence (AI). The EU's planned legislation could be the first to specifically regulate AI."

Microsoft Calls for A.I. Rules to Minimize the Technology’s Risks - "Microsoft endorsed a crop of regulations for artificial intelligence on Thursday, as the company navigates concerns from governments around the world about the risks of the rapidly evolving technology."

Expert Opinions

Embracing the Future: How Chatbots Can Become Great Companions - "Psychologist Abraham Maslow famously posited that human behavior can be classified into a hierarchy of needs, starting with the physiological and ending with self-actualization."

Explainers

All languages are NOT created (tokenized) equal - "Large language models such as ChatGPT process and generate text sequences by first splitting the text into smaller units called tokens. In the image below, each colored block represents a unique token."

---

Copyright © 2023 Skynet Today, All rights reserved.