deepmind

For decades, Shane Legg has anticipated the arrival of “artificial general intelligence” or AGI. 

In other words: an artificial agent that can do all the kinds of cognitive tasks that people can typically do, and possibly more...

Now as the Chief AGI Scientist and a co-founder of Google DeepMind, he stands by that prediction and is calling on the world to prepare. 

This week on How I Built This Lab, Shane’s path to becoming an early AI expert and the work he and his team are doing to prepare for the technological revolution ahead. 

This episode was produced by Sam Paulson with music composed by Ramtin Arablouei. It was edited by John Isabella with research help from Carla Esteves. Our audio engineer was Cena Loffredo.

You can follow HIBT on X & Instagram, and email us at hibt@id.wondery.com.

See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.

This week’s episode is a conversation with Demis Hassabis, the head of Google’s artificial intelligence division. We talk about Google’s latest A.I. models, Gemini and Gemma; the existential risks of artificial intelligence; his timelines for artificial general intelligence; and what he thinks the world will look like post-A.G.I.

Additional listening and reading: 

• A.I. Could Solve Some of Humanity’s Hardest Problems. It Already Has.
• This interview was recorded on Wednesday. Since then, Google has temporarily suspended Gemini’s ability to generate images of humans, following criticism of images the chatbot generated of people of color in Nazi-era uniforms.
• Google Is Giving Away Some of the A.I. That Powers Chatbots

We want to hear from you. Email us at hardfork@nytimes.com. Find “Hard Fork” on YouTube and TikTok.

Our 149th episode with a summary and discussion of last week's big AI news!

Check out our sponsor, the SuperDataScience podcast. You can listen to SDS across all major podcasting platforms (e.g., Spotify, Apple Podcasts, Google Podcasts) plus there’s a video version on YouTube.

Read out our text newsletter and comment on the podcast at https://lastweekin.ai/

Email us your questions and feedback at contact@lastweekin.ai

Some recommended resources for keeping up with AI research:

• Davis Summarizes Papers
• AK’s Substack
• Deep Learning Weekly
• NLP News

Timestamps + links:

• (00:00:00) Intro / Banter
• (00:08:13) Reflections on 2023
• Tools & Apps
  • (00:26:32) Midjourney Version 6: Look Ma, Text!
  • (00:32:09) Baidu's ChatGPT-like Ernie Bot has more than 100 mln users
  • (00:36:05) Android Auto will have Google Assistant summarize your messages with AI
  • (00:37:00) Some of the Samsung Galaxy S24's key AI features just leaked
  • (00:38:33) Microsoft quietly launches dedicated Copilot app for Android
• Applications & Business
  • (00:39:00) Anthropic forecasts more than $850 mln in annualized revenue rate by 2024-end
  • (00:42:25) Anthropic, the OpenAI rival, is in talks to raise $750 million in funding at an $18.4 billion valuation
  • (00:44:10) OpenAI annualized revenue tops $1.6 billion
  • (00:44:46) Nvidia releases slower, less powerful AI chip for China
  • (00:47:30) SMIC is reportedly working on 3nm process technology despite US sanctions
  • (00:50:16) Huawei Files A Patent That Enhances Wafer Alignment And Efficiency, Hinting At The Company’s Self-Built Fabrication Plants
  • (00:52:36) ASML ships first "High NA" lithography system to Intel -statement
• Research & Advancements
  • (00:54:45) Unified-IO 2: Scaling Autoregressive Multimodal Models with Vision, Language, Audio, and Action
  • (00:59:11) Task Contamination: Language Models May Not Be Few-Shot Anymore
  • (01:02:49) Evaluating Language-Model Agents on Realistic Autonomous Tasks
  • (01:08:14) TinyGPT-V: Efficient Multimodal Large Language Model via Small Backbones
  • (01:09:47) Improving Text Embeddings with Large Language Models
  • (01:10:41) DeWave: Discrete EEG Waves Encoding for Brain Dynamics to Text Translation
• Policy & Safety
  • (01:13:23) The Times Sues OpenAI and Microsoft Over A.I. Use of Copyrighted Work
  • (01:18:12) Congress warns science agency over AI grant to tech-linked think tank
  • (01:21:48) Elon Musk’s xAI Jumps on the Bandwagon of Rich Startups 'Benefiting Humanity'
• (01:23:40) Outro

Our 147th episode with a summary and discussion of last week's big AI news!

Correction: Gemini also supports audio

Also check out our sponsor, the SuperDataScience podcast. You can listen to SDS across all major podcasting platforms (e.g., Spotify, Apple Podcasts, Google Podcasts) plus there’s a video version on YouTube.

Read out our text newsletter and comment on the podcast at https://lastweekin.ai/

Email us your questions and feedback at contact@lastweekin.ai

Timestamps + links:

• (00:00:00) Intro/Sponsor Read
• Tools & Apps
  • (00:03:00) Google Just Launched Gemini, Its Long-Awaited Answer to ChatGPT
  • (00:14:15) Google’s best Gemini demo was faked 
  • (00:19:33) Early impressions of Google’s Gemini aren’t great
  • (00:22:30) Google unveils AlphaCode 2, powered by Gemini
  • (00:24:50) Pixel 8 Pro — the first smartphone with AI built in — is now running Gemini Nano
  • (00:25:55) Meta unveils Audiobox, an AI that clones voices and generates ambient sounds
  • (00:29:55) Oops! Elon Musk's Grok AI Caught Plagiarizing OpenAI's ChatGPT
  • (00:33:21) Elon Musk Fans Horrified When His Grok AI Immediately "Goes Woke"
  • (00:34:51) Gmail’s AI-powered spam detection is its biggest security upgrade in years
  • (00:35:55) Augmenting Local AI with Browser Data: Introducing MemoryCache
• Applications & Business
  • (00:37:23) Meta and Microsoft say they will buy AMD’s new AI chip as an alternative to Nvidia’s
  • (00:42:02) China poised to break 5nm barrier — Huawei lists 5nm processor presumably built with SMIC tech, defying U.S. sanctions
  • (00:46:44) OpenAI’s Altman Ouster Was Result of Drawn-Out Tensions
  • (00:50:20) The OpenAI Board Member Who Clashed With Sam Altman Shares Her Side
  • (00:51:46)  Google announces the Cloud TPU v5p, its most powerful AI accelerator yet
  • (00:54:08) Tesla’s Dojo Supercomputer Head Exits in Blow to Efforts
  • (00:56:45) AssemblyAI lands $50M to build and serve AI speech models
  • (00:57:41) Sydney-based generative AI art platform Leonardo.Ai raises $31M
• Projects & Open Source
  • (00:59:20) Mixtral of experts
  • (01:05:50) Meta AI unveils ‘Seamless’ translator for real-time communication across languages
  • (01:07:07) Paving the way to efficient architectures: StripedHyena-7B, open source models offering a glimpse into a world beyond Transformers
  • (01:10:15) Introducing Purple Llama for Safe and Responsible AI Development
• Research & Advancements
  • (01:13:21) WALT is a new AI video tool that creates photorealistic clips from a single image — you have to see it to believe it
  • (01:14:20) Long context prompting for Claude 2.1
  • (01:17:40) Real-World Humanoid Locomotion with Reinforcement Learning
  • (01:19:04) Defending ChatGPT against jailbreak attack via self-reminders
  • (01:21:20) Beyond Human Data: Scaling Self-Training for Problem-Solving with Language Models
  • (01:23:40) Who is leading in AI? An analysis of industry AI research
• Policy & Safety
  • (01:26:41) E.U. Agrees on Landmark Artificial Intelligence Rules
  • (01:33:58) More Trouble Brews For Microsoft As FTC Allegedly Starts Inquiring Into OpenAI Investment
  • (01:39:12) Asking ChatGPT to Repeat Words ‘Forever’ Is Now a Terms of Service Violation
  • (01:41:22)  US in talks with Nvidia about AI chip sales to China
  • (01:44:12) MIT group releases white papers on governance of AI
  • (01:47:25) G7 agrees on first comprehensive guidelines for generative AI
• Synthetic Media & Art
  • (01:49:03) High Court rules that Getty v Stability AI case can proceed
  • (01:52:55) Top Execs at Sports Illustrated's Publisher Fired After AI Debacle
• (01:55:57) Outro

Data privacy has become a major concern for businesses and individuals when it comes to using generative AI tools. What happens to our data? How can we make sure to govern AI properly? Today Katharina Koerner from Tech Diplomacy Network joins us to discuss AI privacy and governance.

Newsletter: Sign up for our free daily newsletter
More on this: Episode Page
Join the discussion: Ask Katharina and Jordan questions about AI and research
Upcoming Episodes: Check out the upcoming Everyday AI Livestream lineup
Website: YourEverydayAI.com
Email The Show: info@youreverydayai.com
Connect with Jordan on LinkedIn

Timestamps:
[00:01:10] Daily AI news
[00:04:20] Understanding AI governance and privacy
[00:09:40] How to balance using AI but keeping privacy
[00:12:45] Privacy-enhancing technologies for secure data sharing
[00:16:00] AI's impact on democracy
[00:20:20] Can AI models be fully transparent?
[00:23:00] Are certain countries ahead in AI governance?

Topics Covered in This Episode:
- Importance of data privacy and concerns about AI tools
- Announcement of Google's new way of watermarking AI images
- Microsoft president's warning about the need for human control in AI
- Release of the enterprise version of ChatGPT by OpenAI
- Introduction of guest expert Katharina Koerner from the Tech Diplomacy Network
- General explanation of AI governance and privacy
- Globally accepted privacy principles and their application to AI systems
- Specific issues related to the origin of data used by large language models
- Differences in regulations around web scraping between the US and Europe

Keywords:
AI, data privacy, AI tools, generative AI tools, AI images, deep mind, Synth ID, watermark, label images, Microsoft president, AI needs human control, tool AI news, AI governance, ChatGPT enterprise, OpenAI, ChatGPT, tech diplomacy network, privacy principles, data quality, data collection limitation, use limitation, transparency, accountability

This Week in Startups is brought to you by…

OpenPhone. Create business phone numbers for you and your team that work through an app on your smartphone or desktop. TWiST listeners can get an extra 20% off any plan for your first 6 months at openphone.com/twist

Crowdbotics. Great ideas can change the world, and Crowdbotics is the fastest way to turn those ideas into code. Get a free scoping session for your next big app idea at crowdbotics.com/twist

Carta now lets you launch and administer SPVs for your syndicate. Share your knowledge, capital, and network to launch your syndicate SPVs through Carta. Get 10% off your first SPV at carta.com/twist with promo code TWIST.

*

Today’s show:

Inflection AI CEO Mustafa Suleyman joins Jason to discuss DeepMind’s influence on AI (1:17), the hardware revolution (26:27), the foundation of Inflection AI (49:45), and much more!

*

Time stamps:

(0:00) Inflection’s CEO Mustafa Suleyman joins Jason

(1:17) Life before DeepMind

(8:34) OpenPhone - Get 20% off your first six months at https://openphone.com/twist

(10:05) DeepMind’s origin story

(15:58) The pitch to Peter Thiel and feedback from early investors

(19:34) DeepMind’s first project

(24:59) Crowdbotics - Get a free scoping session for your next big app idea at crowdbotics.com/twist

(26:27) The hardware revolution and growth in computing

(30:15) DeepMind’s acquisition by Google and strategy for building an elite team

(34:14) Major achievements at Google

(37:49) Carta - Get 10% off your first SPV at https://carta.com/twist with promo code TWIST

(39:22) Google’s cautious approach to releasing an AI product

(44:59) Regulatory issues and fair compensation

(49:45) The foundation of Inflection AI

(59:50) The downsides of AI and Mustafa’s book, The Coming Wave

(1:08:58) Complacency in tech and Mustafa’s thoughts on remote work

*

Follow Mustafa: https://twitter.com/mustafasuleyman

Check out Inflection AI: https://inflection.ai

*

Read LAUNCH Fund 4 Deal Memo: https://www.launch.co/four

Apply for Funding: https://www.launch.co/apply

Buy ANGEL: https://www.angelthebook.com

Great recent interviews: Steve Huffman, Brian Chesky, Aaron Levie, Sophia Amoruso, Reid Hoffman, Frank Slootman, Billy McFarland, PrayingForExits, Jenny Lefcourt

Check out Jason’s suite of newsletters: https://substack.com/@calacanis

*

Follow Jason:

Twitter: https://twitter.com/jason

Instagram: https://www.instagram.com/jason

LinkedIn: https://www.linkedin.com/in/jasoncalacanis

*

Follow TWiST:

Substack: https://twistartups.substack.com

Twitter: https://twitter.com/TWiStartups

YouTube: https://www.youtube.com/thisweekin

*

Subscribe to the Founder University Podcast: https://www.founder.university/podcast

How is AI taking form when it comes to helping businesses in healthcare? Dr. Nadia Boutaoui, Founder of NanoNares Inc,  joins us and shares valuable insights and use cases, discussing how AI can improve operations, enhance patient care, and revolutionize personalized medicine.

Newsletter: Sign-up for our free daily newsletter
More on this: Episode Page
More on this topic in today's newsletter
Join the discussion: Ask Dr. Nadia and Jordan questions about AI in Healthcare
Upcoming Episodes: Check out the upcoming Everyday AI Livestream lineup
Website: YourEverydayAI.com
Email The Show: info@youreverydayai.com
Connect with Jordan on LinkedIn

Timestamps:
[00:00:18] Daily AI news

[00:05:55] How AI improves healthcare

[00:07:30] AI's role in caring for aging people

[00:11:17] Mistrust of AI in healthcare and hallucinations

[00:17:14] How AI helps with chronic diseases, cancer, and new drugs

[00:19:41] AI's role in healthcare customer service and burnout

[00:21:58] Fragmented US health system

Topics Covered in This Episode:
- Caring for aging individuals in their homes
    - Growing issue with the aging baby boomer population in the US.
    - Anticipation of integrating AI technologies for remote healthcare support.
- Future of personalized healthcare
    - Data monitoring and chatbots trained on personal data for treatment.
- Concerns about mistrust of AI and hallucinations
    - Inaccurate information provided by language models.
    - Need to address concerns as a society.
- Using chatbots to answer FAQs and provide personalized responses
    - Segmentation of patient population for targeted communication.
- AI to identify bottlenecks and improve staff utilization
    - Burnout as a significant problem in healthcare.
- Supercomputers for personalized treatments in cancer
    - DeepMind and IBM Watson using large data sets.
    - Collaboration between Regeneron Pharmaceuticals and Geisinger Health System for new drug development.
- Human oversight in advanced data analytics and AI
    - Human validation of recommendations.
- Concerns about data safety and third-party access
    - Importance of complying with HIPAA regulations

Keywords:
technology, healthcare, regulations, burnout, complexity, cost, training, workflows, safety, compliance, HIPAA, adoption, AI, electronic health records, intervention medicine, prevention, asthma, scheduling, specialists, patient care, aging individuals, diabetes, heart disease, centralized data centers, state laws, federal funding, user feedback, personalization, mistrust, hallucinations, chatbots, genomics, electronic medical records, drug development.

Our 91st episode with a summary and discussion of last week's big AI news!

Applications & Business

Meet the DeepMind mafia: These 18 alumni from Google's AI research lab are raising millions for their own startups, from climate to crypto

Nvidia unveils new technology to speed up AI, launches new supercomputer

Kroger and Nvidia partner to ‘reinvent the shopping experience’ with AI and digital twin simulations

Top AI execs, including Richard Socher, launch AIX Ventures

Researchers From Cortical Labs Develop DishBrain: A Neural Network With Biological Neurons

Google AI and UC Berkely Researchers Introduce A Deep Learning Approach Called 'PRIME' That Generates AI Chip Architectures By Drawing From Existing Blueprints And Performance Figures

Seeing an elusive magnetic effect through the lens of machine learning

AI confirms the obvious: The pandemic bummed people out

US-China collaboration in AI papers drops amid ongoing tech war, Stanford report shows

Banks Using AI Are Ripe for Russian Sabotage

Automation will erase 'knowledge jobs' before most blue collar jobs: Future Today Institute CEO

Ukraine uses facial recognition to identify dead Russian soldiers, minister says

Software vendors are pushing "explainable A.I." that often isn't

Ubisoft shows off machine learning tool for modelling animals

Our 87th episode with a summary and discussion of last week's big AI news!

Subscribe: RSS | Apple Podcasts | Spotify | YouTube

Outline:

• (00:00) Intro
• (01:30) AI Arrives for Serious Photo Editing, Not Just Smartphone Snapshots
• (05:20) Listen to an AI voice actor try and flirt with you
• (10:28) DeepMind’s AI can control superheated plasma inside a fusion reactor
• (15:55) Good News About the Carbon Footprint of Machine Learning Training
• (21:36) Who Is Behind QAnon? Linguistic Detectives Find Fingerprints
• (26:18) Facial recognition firm Clearview AI tells investors it is seeking massive expansion beyond law enforcement
• (31:23) This AI Generates Fake New Pokémon, And It's Actually Pretty Good
• (34:01) AI generated this moody, entertaining animated music video
• (37:35) Outro

Our 73rd episode with a summary and discussion of last week's big AI news!

• Applications
  • First AI Pathology Program Approved: Helps Detect Prostate Cancer
  • Adobe Lightroom is getting more powerful with AI-based selection tool 
• Research
  • DeepMind’s AI predicts almost exactly when and where it’s going to rain
  • Summarizing Books with Human Feedback
• Ethics
  • AI’s Islamophobia problem 
  • MIT study finds Tesla drivers become inattentive when Autopilot is activated 
• Fun
  • Amazon’s Astro Is a Mobile Alexa and Cup Holder that Costs $1k

Subscribe: RSS | iTunes | Spotify | YouTube

Find this and more in our text version of this news roundup:  https://lastweekin.ai/p/136

Music: Deliberate Thought, Inspired by Kevin MacLeod (incompetech.com)

Follow the crew:

https://twitter.com/chamath

https://linktr.ee/calacanis

https://twitter.com/DavidSacks

https://twitter.com/friedberg

Follow the pod:

https://twitter.com/theallinpod

https://linktr.ee/allinpodcast

Invest with the Best... ies!:

https://www.thesyndicate.com/allin

Referenced in the show:

NYT - The Rich Kids Who Want to Tear Down Capitalism

https://www.nytimes.com/2020/11/27/style/trust-fund-activism-resouce-generation.html

Coinbase - An upcoming story about Coinbase

https://blog.coinbase.com/upcoming-story-about-coinbase-2012afc25d27

NYT - ‘Tokenized’: Inside Black Workers’ Struggles at the King of Crypto Start-Ups

https://www.nytimes.com/2020/11/27/technology/coinbase-cryptocurrency-black-employees.html

Show Notes:

0:00 Besties intro, fashion talk & the All-In Syndicate

3:47 Insights on Slack being acquired by Salesforce for $27.7B - why did they sell, are there any nitpicks, was this a home run for Salesforce?

19:14 Sacks on going up against Marc Benioff and Salesforce Chatter while at Yammer, importance of holding onto winners as long as possible

27:26 Why is DeepMind's AlphaFold a major breakthrough? When will it begin impacting drug discovery? Dangers of AI at scale

47:07 Republican party misfires, Georgia runoff implications, can Trump pre-pardon his family, Biden's effective strategy so far

59:21 Trump's Section 230 gambit, will Trumpism fade away?

1:05:23 Trust Fund socialists

1:26:14 Coinbase vs. NYT round 2, https://www.thesyndicate.com/allin