• AI Valley
  • Posts
  • ๐Ÿถ UK got the AI source code

๐Ÿถ UK got the AI source code

PLUS: 1.5M Users and 10M Chatbot Test

Good morning. We've got some incredible updates, from the UK government getting the source code of all major AI models to Netflix saving a whopping $1 billion using AI, so letโ€™s begin.

Hereโ€™s what we've got for you today:

  • ๐Ÿ‘จโ€๐Ÿ’ป UK government has the AI source code

  • ๐Ÿค– 1.5M Users and 10M Chatbot Test

  • ๐Ÿ’ฐ Netflix saved $1 billion by using AI

  • ๐Ÿ› ๏ธ 13 New Tools and Resources

Todayโ€™s reading time is 4 minutes.

๐Ÿ‘จโ€๐Ÿ’ป UK government has the AI source code 

So this news is a big one. Big players - Google DeepMind, OpenAI, and Anthropic have agreed to grant the U.K. government full priority access to their AI models for research and safety purposes, according to Prime Minister Rishi Sunak's announcement at London Tech Week.

They said they want to understand the opportunities and risks associated with these AI systems. Sunak expressed enthusiasm for AI's potential to transform sectors like education and healthcare, emphasizing the importance of safety in its development.

Big Plans: The U.K. government aspires to be a global leader in AI safety regulation and plans to host a global summit on AI safety in the fall. Additionally, a Foundation Model Taskforce will receive ยฃ100 million in funding to pioneer research on AI safety and assurance techniques.

Government and big tech working together and sharing the most powerful technology in history. What can possibly go wrong?

A team of Israeli researchers from AI lab AI21 recently conducted a study involving the largest-ever Turing-style test, involving 1.5 million users and over 10 million chat sessions.

The Goal: They wanted to determine if participants could distinguish between human and AI chat partners. The study found that human users correctly identified AI chatbots only 60% of the time, just slightly better than chance.

ChatBots: The researchers created AI chatbots using various language models, including GPT-4, Cohere, and AI21's Jurassic-2. The bots were given distinct personalities and traits to enhance their believability.

Strategies: People tried all sorts of strategies to figure out who was who, like discussing current events or asking questions in foreign languages, but they were not effective. However, participants who made complex or illegal requests correctly identified bots 65% of the time.

Runway(backed by Google) finally released its newest model, Gen-2, which generates videos from text prompts or images. No more waitlist. This could be the next big thing in AI, except it's less "Hollywood blockbusterโ€.

The Problem: Testing the model was a bit like asking your grandpa to take up beatboxing. The videos ran at a framerate so slow that they had a grainy-looking quality. The video feels so weird to watch sometimes and has viral meme potential because of its weirdness.

The Big Picture: But this is just the beginning of AI-video models. Remember how Midjourney (text-to-image) used to generate images a year ago? And look at it now:

V1 was released 1 year ago

So this may happen with videos in less than one year (all hail exponential growth).

Here are 12 incredible videos generated using Gen2. Have a look

Zoom Out: As AI like Runway's Gen-2 evolves, it's easy to imagine an individual sitting in their room, armed with nothing but a creative mind and a basic computer, becoming a video producer. No high-end cameras, no expensive lighting, no location scouting, and no Hollywood-sized budgets.

๐Ÿ“Š STATS & TRENDS

Netflix saved $1 billion by using machine learning ๐Ÿ’ฐ

In 2017, Forbes reported that Netflix harnessed machine learning, leading to $1 billion in savings.

This was driven by research indicating users abandon a search within 90 seconds. Netflix's response was an algorithm offering personalized recommendations, a tactic also adopted by Amazon Prime and Hulu.

Full research was done by Netflix - here

๐Ÿ—ž๏ธ EXTRA ROUNDUPS

๐ŸŒ NASA warns of an โ€œInternet Apocalypseโ€ that would disable the Internet for months.

๐Ÿšซ Thousands of Reddit communities will be inaccessible on Monday in protest at how the site is being run.

๐Ÿ’ฐ Salesforce pledges to invest $500M in generative AI startups.

๐Ÿฆ TOOLS & TREATS

  • METAโ€™s MusicGen - It produces high-quality music while being conditioned on text descriptions or melodic features.

  • Audio Pen - An app that converts your voice notes into concisely summarized text.

  • Altered AI - Change your voice to any of our custom-curated voices for professional performances.

  • Dummie AI - Turn your videos into engaging shorts that outperform, no editing needed.

  • Chat2Course - The platform offers a personal course builder and AI tutor, allowing you to customize your learning experience.

  • PainFinder - Ask questions related to the pain points users encounter when using a particular software or software category.

  • Sheet Plus - Generate Google Sheets & Excel formulas from text, convert formulas into simple explanations, debug formulas, and more.

  • YourSearch - An AI-powered search engine designed to facilitate personalized internet research directly in your favorite messenger

Refer 1 friend to get all the links to our previous issues (using this link)

๐Ÿ˜† MEME OF THE DAY

Credit to my friend - Gaut

TWEET OF THE DAY ๐Ÿ”ฅ

All the major AI developments that have happened so far in 2023:

Learn new things and generate new ideas using this tool:

Thatโ€™s all for today, I hope you enjoyed it.

Written with ๐Ÿ’š and Thank you, everyone, for being part of our journey, it means a lot to me and my team ๐Ÿถ

AWESOME PERSON OF THE DAY ๐Ÿ’š

What did you think of today's newsletter:

Your feedback helps to create better content for you.

Login or Subscribe to participate in polls.

If you have any inquiries or questions regarding anything, please feel free to dm me on Twitter: @heyBarsee