Meta's LlaMa 2 is now Commercially Available
PLUS: Google Deepmind's AI for Diagnosis, OpenAI + Local News Org, Bing Chat becomes Multimodal
Hey there 👋
Meta flooded the internet last night with its new opensource model Lllama 2, making it available for commercial use as well. Clearly, Meta is fearless about who uses it, however, a lot of measures have been taken to ensure safety. But, jailbreaks are never impossible! The intriguing part of this was Meta’s partnership with Microsoft for releasing Llama 2. Microsoft’s alliance with OpenAI which releases advanced models but doesn’t opensource them, and now Meta. Are we seeing the smart leader here?
Further, OpenAI made another interesting move by joining hands with AJP to embed AI in local news reporting, indicating their efforts to steer clear of copyright infringement lawsuits.
This issue covers:
Latest Developments 🌍
Tools of the Trade ⚒️
Trending AI Jobs 💼
Prompt of the Day 📜
Hot Takes 🔥
AI Meme of the Day 🤡
Read time: 3 mins
Latest Developments 🌍
Meta’s Llama 2: Llama 2 is a collection of pretrained fine-tuned LLMs ranging from 7B to 70B parameters. Here are the key features:
Available in three sizes 7B, 13B and 70B parameters,
Trained on 2 trillion tokens (40% more data than Llama 1), and over 1M human annotations,
Fine-tuned using SFT and RLHF to align with human values, red-teaming used to ensure more safety,
4k context length,
Outperforms other opensource models including MPT and Falcon in various benchmarks for coding, reasoning, language and other tasks,
Performance close to GPT 3.5 on language and maths benchmarks, but loses with a significant gap in coding,
Available on Microsoft Azure AI, AWS, Hugging Face, and other providers.
Llama 2 on Mobile: Meta and Qualcomm are collaborating to enable Meta's Llama 2 to run on Qualcomm chips in phones and PCs from 2024.
Bing Chat becomes Multimodal: Microsoft made the following announcements:
Enterprise version for Bing Chat
Microsoft 365 Copilot priced at $30 per user per month
Multimodal capabilities via Visual Search in Bing Chat for web search and QA.
Google's Latest Models on Vertex: Google Cloud is expanding the availability of enterprise-ready foundation models on Vertex AI: Imagen (image generation, editing and visual QA), PaLM 2 for Chat, Codey (code generation, completion and code chat), and Chirp (multilingual speech AI).
Google Deepmind’s AI for Diagnosis: CoDoc interprets medical images and gives a predictive confidence score to determine when to rely on predictive AI and when to defer to a clinician for better interpretation.
Local News and OpenAI: OpenAI is committing $5+ million (free credits) to American Journalism Project, for creating a technology studio where the local news companies will apply OpenAI’s technology to enhance journalism.
OpenAI’s reign on Replit: AI projects on Replit experienced an explosive 34-fold YoY growth in the first half of 2023, with OpenAI dominating over 80% of distinct projects.
Authors Fight Back: Thousands of authors have signed an open letter urging companies to stop using their copyrighted work to train AI models without their permission or compensation, citing potential flooding of mediocre AI-generated books, harming their livelihood and the publishing industry.
'World’s first mass-produced' humanoid robot: Shanghai-based firm, Fourier Intelligence, is developing the world's first mass-produced humanoid robot, the GR-1, to serve as a caregiver, therapy assistant, and companion for the elderly.
AI for Insurance Inspection: Tractable secures $65M in Series E funding for its AI computer vision technology used in insurance assessments for property and car damage.
Challenging Google's Dominance: South Korea's Naver plans to release its AI model soon to counter Google's impact on its core search engine business. It will collaborate with Samsung to develop chips for data processing servers.
G/O's Controversial AI Push: Despite the initial backlash, Gizmodo Media plans to use more AI-written articles, potentially replacing humans with low-value content.
UNSC’s First AI Meeting: While the member countries acknowledged AI to be a ‘double-edged sword’, Russia questioned the very discussion by the UN Security Council about AI.
Tools of the Trade ⚒️
Llama Chat: Chat with Meta AI’s new model Lllama 2 on Perplexity AI platform.
XO Analytics: Get insights for your entire e-commerce business along with recommendations.
Nowy: AI visual travel planner that lets you discover hidden gems, plan itinerary, connect with fellow travelers, and make instant bookings.
Jackrabbit Ops: Automate handling RFQs, identifying customers with intent to buy, and booking meetings.
Juno: Your Jupyter copilot that writes, edits, and automatically debugs your code with you as you go.
Page One: AI-powered news platform tailored to your preferences, lets you keep track of important stories.
GenWorlds: Create your own multi-agent world with each AI agent acting independently with specific personalities, memories and expertise.
😍 Enjoying so far, TWEET NOW to share with your friends!
Trending AI Jobs 💼
Meta: Partner Engineer, Generative AI |Menlo Park, CA (Apply Now)
Duolingo: Senior AI Research Engineer | New York, US (Apply Now)
Anthropic: Research Engineer, Product | San Francisco, CA (Apply Now)
Prompt of the Day 📜
AI as Job Interviewer
Imagine that you are a job interviewer speaking with a candidate for a [Job Title] position at [Your Company]. Ask relevant questions about the candidate's skills, experience, and qualifications, as well as questions that explore their problem-solving abilities, team collaboration, and alignment with the company's culture and values.
Hot Takes 🔥
Meta's openness and lack of fear of AGI will, against the odds, win them the generative AI race. ~Alberto Romero
AGI is the riddle of history solved and it knows itself to be this solution. ~Bojan Tunguz
Meme of the Day 🤡
That’s all for today!
See you tomorrow with more such AI-filled content. Don’t forget to subscribe and give your feedback below 👇
Real-time AI Updates 🚨
⚡️ Follow me on Twitter @Saboo_Shubham for lightning-fast AI updates and never miss what’s trending!!
PS: I curate this AI newsletter everyday for FREE, your support is what keeps me going. If you find value in what you read, share it with your friends by clicking the share button below!
When they say "safety", they mean "censorship". When they say "align with human values", they mean "align with politically correct values". I want the uncensored version that will provide whatever information it has on topic I ask it about and will fairly output the pros and cons from all sides any controversial or politically charged topics without biasing its response toward either side.