- Work With AI
- Posts
- ChatGPT Introduces Plugins and Microsoft Researchers See a Spark of AGI in GPT-4
ChatGPT Introduces Plugins and Microsoft Researchers See a Spark of AGI in GPT-4
March 24, 2023
⚡ Today’s Highlights
📰 News: ChatGPT introduces plugins and researchers think GPT-4 could be an early version of AGI
💰 Funding: Character.ai, Vital, Workera, OneThird
🦾 Tools: A search engine with intelligent memory and an AI for turning ideas into videos with compelling stories and visuals
📅 Events: Microsoft Secure and MemCon are both happening next Tuesday (March 28)!
📰 Today's Top Stories
(4 min read) (Source: OpenAI)
TLDR: OpenAI launched plugins for ChatGPT, granting it access to third-party knowledge sources and databases.
The first-party web-browsing plugin allows ChatGPT to draw data from around the web using the Bing search API to retrieve content from the web and cite its sources in responses.
The first plugins have been created by Expedia, FiscalNote, Instacart, KAYAK, Klarna, Milo, OpenTable, Shopify, Slack, Speak, Wolfram, and Zapier.
Access is currently limited to a small number of developers and ChatGPT Plus subscribers.
To encourage plugin creation, OpenAI has open-sourced a retrieval plugin that lets ChatGPT access snippets of documents from data sources through natural language questions.
(30 min read) (Source: Microsoft)
TLDR: Researchers at Microsoft published research showing that GPT-4 could be viewed as a very rudimentary and incomplete version of Artificial General Intelligence (AGI).
OpenAI's latest model, GPT-4, was trained on an unprecedented scale of compute and data, making it part of a new cohort of large language models with more general intelligence.
GPT-4 can solve difficult tasks spanning mathematics, coding, vision, medicine, law, psychology, and more without special prompting, with performance often surpassing previous models like ChatGPT.
Microsoft researchers contend that GPT-4's breadth and depth of capabilities allow it to be reasonably viewed as an early, incomplete version of an artificial general intelligence (AGI) system.
Research emphasizes discovering limitations and challenges ahead for advancing towards deeper and more comprehensive versions of AGI, including the possible need for a new paradigm beyond next-word prediction.
The Big Picture: Artificial General Intelligence is generally defined as an AI system with an intelligence equal to humans. It is sometimes referred to as “Strong AI”, an AI with self-aware consciousness that has the ability to solve problems, learn, and plan for the future, though Strong AI is usually reserved for AI with consciousness where AGI may just have human-level intelligence. AGI is considered by most to be the pinnacle of AI research and the goal most AI companies are working towards.
(5 min read) (Source: TechCrunch)
TLDR: Canva unveiled new features and tools during its Canva Create virtual event, including a suite of AI-powered tools and a new Brand Hub.
Assistant lets users search for design elements, provides quick access to features, and gives recommendations on graphics and styles that match your existing design.
Magic Write, the platform’s AI-powered copywriting assistant, is now accessible within all Canva project types and available in 18 languages.
Other new tools include "Magic Design" to get a curated selection of personalized templates, "Magic Edit" to add or replace anything in an image, "Magic Eraser" to clean up unwanted details, "Beat Sync" to match video footage to a soundtrack of your choice, and "Translate" to automatically translate text in designs to over 100 different languages.
(4 min read) (Source: Reuters)
TLDR: France's National Assembly has approved the use of AI video surveillance to enhance security during the 2024 Paris Olympics.
The technology can detect predetermined events and crowd surges, but it is criticized by civil rights groups who believe it is a threat to civil liberties.
If it is formally adopted, France will become the first country in the European Union to legalize AI-powered surveillance.
💰 Funding Alerts
Character.AI is a conversational AI technology company that uses large language models and deep learning algorithms to provide personalized artificial intelligence experiences. The company just officially announced that they raised $150 million in a Series A funding round led by Andreessen Horowitz (Sarah Wang, General Partner), with participation from previous investors such as Nat Friedman, Elad Gil, SV Angel, and A Capital. The company plans to use the funds to expand its computing abilities, develop new features, and grow its engineering team.
Vital is an AI-driven digital health company that uses NLP and AI to provide patients with educational content during their emergency department and inpatient visits. The company recently raised $24.7 million in Series B funding, bringing its total funding to over $40 million. The funding was led by Transformation Capital, with participation from Threshold Ventures and Aaron Patzer. Vital plans to use the funds to expand its team and launch new features to enhance clinical decision support, care coordination, and patient experience.
Workera is a skill development platform that diagnoses an organization's skills to empower leaders to make better talent development decisions. The platform uses deep learning to deliver personalized assessments and learning plans for employees. Workera raised $23.5 million in Series B funding, led by Jump Capital, with participation from NEA, Owl Ventures, AI Fund, and Sozo Ventures. The company plans to increase its comprehensive ontology of over 7,000 skills to include soft skills assessments, such as leadership and management, and help employees upskill faster than ever by creating skills-based career pathways.
OneThird is an AI-powered food tech company that uses near-infrared scanning technology to accurately predict the shelf life of fresh produce and minimize food waste. They have raised €2.75 million euros ($3 million USD) in a late-seed funding round, led by Pymwymic with participation from prior investors, Halma Ventures Limited, SHIFT Invest and Oost NL. The company plans to use the new funding to expand its technology and operations to new markets in North America and Europe to tackle the $1 trillion lost annually worldwide to food waste due to spoilage.
🦾 Trending Tools
Rewind is a memory search engine that captures everything seen, said, or heard, making it searchable through ChatGPT for Me, powered by OpenAI's GPT-4. Privacy is prioritized with local storage, and only text is sent to OpenAI for generating answers. Answers include direct links to sources for verification.
Powered by GPT-4 and Whisper, VoicePen automatically generates engaging blog posts from audio and video files. Upload any file, including YouTube clips, podcasts, and webinars, and receive a transcription and four blog post variations.
ChaturGPT is an AI-powered PDF reader that lets users ask questions and receive quick answers without having to read through the entire document. Ideal for students, teachers, and professionals.
Typedesk's new app integrates ChatGPT by OpenAI to create dynamic prompts for all apps and websites, using a single shortcut. Its user-friendly interface and Quick Access window make content creation fast and intuitive. Try it out with a 30-day free trial and share templates with your team!
Visla's new Idea to Video feature allows users to quickly generate videos with compelling stories and visuals on both mobile and desktop. Simply input a few words and the AI will write a story, match relevant stock footage, and sync AI-generated voiceovers.
👀 More Reading
The A.I. Chatbots Have Arrived. Time to Talk to Your Kids. (The New York Times)
3 ways to center humans in your company’s artificial intelligence efforts (MIT Sloan)
Tech guru Jaron Lanier: ‘The danger isn’t that AI destroys us. It’s that it drives us insane’ (The Guardian)
WGA Would Allow Artificial Intelligence in Scriptwriting, as Long as Writers Maintain Credit (Variety)
🌎 Fresh Content
Special edition coming in hot! Today we’re doing a showcase of Canva’s AI text-to-image generator. We’ve put together a few different examples with the prompts used to generate them.
1. It seems like it’s doing pretty well with hands so far, a major issue with early Midjourney models. This first one seems fairly ordinary. The second one is a little… unusual. (Prompt: “shaking hands”)
2. Finer details like the texture of this dog’s fur or the waves in the water seem to look decent as well. Not as high definition as Midjourney V5 but an incredible first version. (Prompt: a golden retriever running on the beach)
3. It still has a long way to go with human faces but did well with a popular monument such as the Eiffel Tower. (Prompt: a group of friends taking a selfie in front of the Eiffel Tower)
4. It seemed to get confused with more unusual requests. Are we the aliens? 🤔 (Prompt: an alien meeting an astronaut)
🧠 Resources
Yufeng Guo of Cloud AI Adventures explains machine learning from the ground up, using concrete examples. In the second video, he walks us through the 7 steps required to create effective machine learning models.
📚 Research
Artificial intelligence predicts genetics of cancerous brain tumors in under 90 seconds (MedicalXpress)
The universe is getting a weigh-in thanks to AI (Popular Science)
📅 Upcoming Events
Microsoft Secure (March 28, 2023 @ 8:30 AM PDT / Virtual) presents conversations on the state of the industry between Microsoft leaders helping to deliver the products security teams use daily. Be among the first to see what an AI-powered future means for cybersecurity, and learn new comprehensive strategies to help you protect more with less. Register today to secure your spot.
MemCon (March 28-29, 2023 @ The Computer History Museum, Mountain View, CA, USA) is the first memory event focused on end users and systems. MemCon empowers engineers and architects working on memory-constrained problems with peer insights from case studies in genomics, AIML, datacenter, HPC, computational fluid dynamics, in-memory databases, and enterprise knowledge graphs. If you are working on a memory-bound problem, come and unpack it with your peers, meet enterprise data experts, and get your hands on tech demos at the same time. Register your interest today.
💼 Jobs
Coda: Software Engineer, AI/ML
Seattle, San Francisco, and Mountain View, USA (Remote available)
$125,000 - $250,000 USD
Oura: Machine Learning Data Scientist
Remote, USA
$140,000 - $180,000 USD
Center for AI Safety: Machine Learning Writer
San Francisco, CA, USA
$110,000 - $150,000 USD
Modern Intelligence: AI Research Assistant
Austin, TX, USA
$65,000 - $75,000 USD
C3.ai: AI Solution Director, Healthcare
Houston, TX; Redwood City, CA; New York, NY; Chicago, IL; Tysons, VA, USA
$184,000 - $216,000 USD
That's all for today's email! If you want more please follow us at the social channels linked below, or check out our website!
How'd you like today's email? |
Share our newsletter: If you like our work please share/forward this email with your friends, colleagues, and family. It's the best way to support us!
If this email was forwarded to you please sign up here to continue receiving them.
Want your content, product, jobs, or event featured in our newsletter? Reply to this email with the details, and our team will reach out to you.
Do you use AI for work? Tell us how, and you could be featured in our newsletter!
Check out our website for more resources, including a list of AI investors, products, events, and twitter follows.
For an archive of all our posts, click here.
We'd love to hear from you! You can always leave us comments or feedback by replying to this email!