OpenAI chases Her
ChatGPT left the textbox and where AI is leading society.
Tom and I recorded an episode of The Retort on OpenAI’s culture shift with this announcement — it’ll be out Friday. Subscribe if that sounds interesting!
Latest OpenAI Articles
Chaos and tension at OpenAI
Safety seems to be taking a back seat
Aaron Turner
•
1d
So many questions! (A) Should the public be worried? -- The public should be worried (in particular) about any AI system that is deployed at scale, so whatever the major AI labs such as OpenAI are doing should be top of this list. All contemporary AI (including the major LLMs such as ChatGPT) is at best only minimally aligned with human values. BY DEFINITION, a minimally-aligned AI system will inexorably inflict societal harm (in parallel with any benefits) when deployed. Therefore a minimally-aligned AI system that is deployed at global scale (such as ChatGPT) will necessarily inflict societal harm at global scale (although it's difficult to predict in advance exactly what form that harm will take, we will doubtless gradually find out over the next 10-20 years). The only good news in this regard is that contemporary LLMs are simply too dumb by themselves to represent a significant x-risk. (B) Will the (new) board at OpenAI take note? -- Internally, for sure; these are not stupid people. (C) Will they do anything to address the situation? -- Only if they perceive it to be a problem relative to their (highly rationalised) objectives. It may take external force, such as robust legislation, to significantly temper their (and thus the company's) behaviour (ditto re the other AI labs of course). (D) Will OpenAI’s status as a nonprofit remain in good standing? -- In my assessment, OpenAI is already firmly under the control of profit-motivated interests. It's entirely possible however that key people within the company, even board members, have not quite worked this out yet. (E) Will [this] help Elon Musk’s case? -- Quite possibly. (F) Does Sam care? -- I believe he cares deeply. I also believe HE believes he's doing the right thing, which (when Kool-Aid is involved) is not necessarily the same as actually doing the right thing. (G) Is this what he wanted? -- I suspect not, but HE needs to work this out (ditto re the rest of the board). (H) Is OpenAI’s push to commercialization coming at the expense of AI safety? -- 10,000%.
Joy in HK
•
1d
One has to look long and hard with a magnifying glass to find examples of a profitable business willing to offset revenue for safety concerns. Ford Pinto, anyone?
John Schulman (OpenAI Cofounder) - Reasoning, RLHF, & Plan for 2027 AGI
how posttraining tames the shoggoth, and the nature of the progress to come...
Chatted with John Schulman (cofounder of OpenAI and led ChatGPT creation) how posttraining tames the shoggoth, and the nature of the progress to come... Watch on YouTube. Listen on Apple Podcasts, Spotify, or any other podcast platform. Read the full transcript
I enjoyed the interview, good stuff 👍
What I Read This Week...
Google's Deepmind releases a new biology prediction tool, Apple is finalizing a deal with OpenAI, and more than a third of 18-24 year-olds reported no income in 2022
Watch All-In E178 Read our latest deep dive into semiconductors Caught My Eye… Google’s DeepMind has released an improved version of its biology prediction tool AlphaFold. While Google’s previous model amazed the research community with its ability to predict protein structures, Google’s latest iteration can predict the structures and interactions of nearl…
Shocking stat about Gen Z. We all need to mentor them to be victors not victims. These charts show that the DEI/ESG administrative state in education and healthcare is making us all poorer, sicker, and dumber: https://yuribezmenov.substack.com/p/fire-dei-esg-hr-commissar-administrative-bloat
Great list, do you think the drop in the young workforce is due to the polarisation of 'get rich quick' and the 'easy' money that can be made online?
OpenAI Rules the Changes But Meta Changes the Rules
An analysis on Meta’s master plan and OpenAI’s masterpiece
A blog about AI that’s actually about people Meta has put the entire AI startup ecosystem against the ropes. They’ve released the two smaller versions of the Llama 3 family (8B and 70B-parameter dense models) and have given us a glimpse at the large version, a 405B dense model that although still training, is already showing
Good job doing your homework, as this allows you to make some very good quality contracture. I’m personally hoping that GPT4 will eventually become free because my business model will be greatly assisted by that development. I’m not sure what the odds are, but of course, if we have the momentum of progress on our side, then it shouldn’t take forever.
Gracias. Muy interesante.
OpenAI GPT-4o: The New Best AI Model in the World. Like in the Movies. For Free
Everything you need to know (so far)
A blog about AI that’s actually about people OpenAI has delivered. In just a brief 25-minute live event they’ve changed the landscape completely. Here’s the image that best reflects why:
Kudos for the quick turnaround on this in-depth piece about GPT-4o. It's impressive how you've captured the essence of this release, highlighting not just the technological advancements, but also the economic and societal implications, with such speed and clarity. Your ability to distill complex developments into an engaging narrative stands out. Once again.
Your crystal ball prediction from this last Saturday was fantastic Alberto! Now a couple of dumb questions: 1). Will this mean that everyone can access the GPT store via either 4 or 4o? 2) Will this multimodality change the very nature of the tools that we can create on the GPT store? The first question has great bearing on my new leaf fledgling business, which is to tap the power of custom GPTs. I always worried about perspective clients barking at the $20 a month. Is that now a done deal?
Brace for Impact: Here Comes the "Cram Down"
Upcoming Edtech Happy Hour Events, ASU+GSV 2024 Session Overviews, US Newspapers Sue OpenAI, Coursera and Chegg Stock Down, and more!
Brace for Impact: Here Comes the “Cram Down” By Ben Kornell
Ben - this is so insightful and so true. Twain said "history doesn't repeat itself, but it often rhymes". I've lived through three of these cycles now - the S&L crisis in '90-93, the dot com and telecom winter from '01 to '04, and then Global Financial Crisis from '08-'11. Every time we go through the same cycle. When a bubble bursts, during the first year people believe that a recovery is right around the corner. It'll be fine! The second year, they realize this may take a while longer and that they need to start cutting costs to extend the runway and avoid exposing themselves to "market pricing discovery". When they run out of moves, they reach the capitulation stage and that's when the dreaded "inside down round" happens. People start to read the deal docs and understand how weighted average anti-dilution provisions really work, what discounts on notes and SAFEs really do to founder economics, and how pay to play provisions work. It's ugly. The companies that get through this phase quickly, or even better proactively in the first two years, are well positioned to be acquirors of both market share and weaker competitors. These cycles typically last 4 years and we're about 18-24 months into this one.
I'm very optimistic about the future. We're seeing strong revenue growth in our portfolio and the long term trends underlying the digitation of education and alternative ways to upskill the workforce are very much intact. It just takes time, but anyone who's been around education for a long time knows that everything takes time in our business.
OpenAI Just Unveiled The Best AI Model: GPT-4o
Here's why everyone is talking about GPT-4o.
OpenAI finally unveiled GPT-4o! This new model offers real-time multimodal capabilities in audio, vision, and text, but now with significant enhancements. It’s free to use, which shows a strategy similar to that used with GPT 3.5—aimed at attracting new users and further scaling model training.
it seems we get our minds blown by AI every few months ! .. one step closer to AGI which is going to be a monster disruption to every industry
498: Constellation Software Annual Meeting & Q1, OpenAI GPT-4o + Apple, Databricks, Intel + Apollo, Google AlphaFold 3, and Solar Storm
"Having fun is a habit."
Great post. Love the quotes from Mark L!
Hot take on OpenAI’s new GPT-4o
GPT-4o hot take: • The speech synthesis is terrific, reminds me of Google Duplex (which never took off). but • If OpenAI had GPT-5, they have would shown it. • They don’t have GPT-5 after 14 months of trying. • The most important figure in the blogpost is attached below. And the most important thing about the figure is that 4o is not a lot different from Tur…
As someone who builds on top of these API's, I think you're underestimating how big of a deal the drastically reduced latency is. Yes the evals look like diminishing returns but the latency and cost improvements are drastic.
Spot on. The ways in which it becomes more convincing that there is actual understanding seem to outpace actual progress on understanding.
In the meantime Sam discusses UBC when GPT-7 arrives. Which is all too much messianic prophet for my taste.
Human intelligence by the way is also amazing as well as often pretty dumb, so who are we to point fingers?
OpenAI fluffs its lines as model collapse makes a mockery of its flagship
Leadership in AI is back up for grabs as Chat GPT-4o fails to advance, sending Microsoft, Google, and Apple, scrambling to gain ground...
Before I get to the big news of the day out of OpenAI, I have a favour to ask. I publish most of my insights for free, because I believe the facts in them are important enough to need as large an audience as possible, but my work isn’t free. Future Media is a reader-supported publication. To receive new posts and support my work, consider becoming a free…
Last Week in AI #269: Better evals for multimodal AI, new OpenAI lawsuits, Meta's AI ads tool troubles, AI startups focus on enterprise, and more!
Reka AI releases Vibe-Eval, 8 US newspapers sue OpenAI, Meta's AI ads tool's overspending problem, AI startups are pivoting to enterprise customers
Top News Vibe-Eval: A new open and hard evaluation suite for measuring progress of multimodal language models Reka AI introduces Vibe-Eval, a new evaluation suite designed to measure the progress of multimodal language models. Researchers from the company have created a set of challenging prompts to test the capabilities of these models, particularly focu…
🔮 Coordinating the energy transition; AI’s hiring bias; new molecules; urban recycling, teachers, and music taste ++ #473
Hi, I’m Azeem Azhar. In this week’s edition, we explore the coordination challenges of the energy transition. And in the rest of today’s issue: Need to know: LLM’s hiring bias GPT-3.5 demonstrated stereotypical biases when evaluating resumes for various occupations.
Love the mention energy, how it relates to climate change.
Unfortunately, the fossil fuel subsidy has many dimensions and its roots run deep, making it difficult to make meaningful progress. From OPEC to oil drilling, gas stations, and trucks, many individuals depend on the #fossilfuel industry for their livelihoods. And as the saying goes - “if your livelihood depends on not understanding something, you won’t”
That's why instead of demonizing those who work in fossil fuel industry, we need to understand what’s their challenge. We need to set our differences aside and work together to find a path forward. We have no time to waste. I'm asking this community for help.
If you or someone you know is working towards creating economic opportunities for those in the fossil fuel industry, please connect with me. I'm eager to offer my problem-solving and communication skills to accelerate progress towards a green future.
Although I'm not an expert in sustainability, I'm committed to understanding the problem better. If you're a sustainability expert and would like to help me think through the problem, please reach out.
The Sam Altman Playbook
Fear, The Denial of Uncertainties, and Hype
How do you convince the world that your ideas and business might ultimately be worth $7 trillion dollars? Partly by getting some great results, partly by speculating about unlimited potential, and partly by downplaying and ignoring inconvenient truths.
Sam is a pseudo-philosopher in a world that has forgotten how to think critically. Thanks for this. Best Gary thus far. Look forward to more.
“all of this has happened before. all of this will happen again.”
Yep, I’ve been there. Working on AI research in 1990-1991, just before the second AI Winter.
Met Gala Not Dead, But Decaying
The Gala itself has a new theme, a new carpet design, and varied guests each year, but has become rather predictable.
Before the Met Gala Monday night, Anna Wintour apologized for “confusion” over the theme. The Costume Institute exhibition at the Metropolitan Museum of Art that the Gala opens and raises money for is Sleeping Beauties: Reawakening Fashion, but the dress code was “The Garden of Time.” Anna said on
It all seemed so joyless.
I wonder if part of the reason is that that the event has outgrown itself in a specific kind of way. When you compare it to the Oscars, for instance, there is a point to it all (the actual awards) that the public is part of, because they can view the entire event. The Met Gala just feels so disconnected now because you know the vast majority of attendees don't care about museums (some of them probably don't even know the difference between the Met and MoMA, I'm guessing) or even fully understand why they are there. So it all just feels kind of fake, and the public isn't allowed inside for the party, so it winds up feeling very hollow, despite the absurd star power on the red carpet.
When it was more of a society event, there was a perception of authenticity, because those attending seemed to have a true interest in the institution of the Met. It felt like more of a NYC-specific type of thing that was more closely connected to the museum. I am not sure if any of this made sense, it's kind of word salad, but tldr: the gala is too big, too corporate, and too phony-feeling now to be relevant.
OpenAI's GPT-4o and Partnership with Apple 🍎
Mira Murati was magical, not sure about all of these accessibility upgrades.
Audio Introduction 🎧 0:54 Hello Everyone, OpenAI had a Spring Update event the day before Google I/O begins, today as I write this on May 13th, 2024. It comes at an awkward time for Google, that lost market share as Apple’s talks with OpenAI have gone well to bring GPT-4o’s Voice to presumably, the upcoming iOS 18 and iPhones of 2025.
Definitely waiting for Siris new update, as well as integration into some big time video games and VR software. Then this takes off to a new level, albeit more niche than these general purpose model, but way more immersive
That’s a very comprehensive take on the Michael. As you said I am not sure whether OpenAI will get the voice assistant right but I am satisfied for one reason: Ability to solve easy math problems.
This is significant because basically what differs AGI from the current models is next-token-prediction.
Current models lack this, if the question wasn’t included in their training data they simply can’t solve it. I am not telling it can’t provide an answer, it will and these answers will sometimes be true, but it won’t get that answer by applying logic predicting the outcome.
The questions illustrated were too basic so they could already be included in the training data. But if they were not included in the data as a block and the mode can solve a basic math problem like 1+2+7+3+5+12+33+17=? then AGI is basically just building on this.
495: XPEL and Anchoring Bias, Apple + OpenAI & Google, Global Coal Fleet, Helium, Sean Feeney, Music, and The Sopranos
"Great writing is precious"
"What if this was more common? Great writing is precious enough that I wish we had multiple interpretations of most great works. It would be a great way to see the evolution of artists."
Yes! Mark Twain on Jane Austen is a good (read as: hilarious) place to start: In his extensive correspondence with fellow author and critic William Dean Howells, Mark Twain seemed to enjoy venting his literary spleen on Jane Austen precisely because he knew her to be Howells’ favorite author, In 1909 Twain wrote that “Jane Austin” [sic] was “entirely impossible” and that he could not read her prose even if paid a salary to do so. Howells notes in My Mark Twain (1910) that in fiction Twain “had certain distinct loathings; there were certain authors whose names he seemed not so much to pronounce as to spew out of his mouth...
Rather than pitying Twain when he was sick, Howells threatened to come and read Pride and Prejudice to him.
Twain marveled that Austen had been allowed to die a natural death rather than face execution for her literary crimes. “Her books madden me so that I can’t conceal my frenzy,” Twain observed, apparently viewing an Austen novel as a book which “once you put it down you simply can’t pick it up.” ... In a letter to Joseph Twichell in 1898, Twain fumed, “I have to stop every time I begin. Everytime I read “Pride and Prejudice” I want to dig her up and beat her over the skull with her own shin-bone.” From: https://www.vqronline.org/essay/barkeeper-entering-kingdom-heaven-did-mark-twain-really-hate-jane-austen
Last Week in AI #268: Gen AI for gene editing, Moderna partners with OpenAI, model releases from Microsoft and Snowflake, and more!
Gen AI used to generate new gene editors like CRISPR, Moderna's internal ChatGPTs, Microsoft releases Phi-3-mini LLM that can run on a phone, Snowflake open sources enterprise LLM
Top News Generative A.I. Arrives in the Gene Editing World of CRISPR Generative AI, which have already revolutionized areas such as art and programming, are now making significant strides in biotechnology. A new A.I. system developed by the Berkeley-based startup Profluent has been designed to create blueprints for novel gene editors by employing methods …
Last Week in AI #270: DeepMind releases AlphaFold 3, Marines test robot dogs with rifles, DeepSeek-V2, TikTok to label AI-generated content, and more!
AlphaFold 3 can predict DNA/RNA and more, Marine Spec Ops test weaponized quadrupeds from Onyx Industries, DeepSeek-V2 improves LLM efficiency, TikTok will label all AI-made content
Top News Google DeepMind’s new AlphaFold can model a much larger slice of biological life Google's DeepMind has unveiled AlphaFold 3, an advanced version of its protein structure prediction tool, which can now predict the structures of DNA, RNA, and essential drug discovery molecules like ligands. This new tool provides a more detailed and dynamic view of…
What to Make of GPT-4o
The latest OpenAI algorithm launched with great promise. What are the impacts of the latest GPT algorithm?
Is the GPT-4o algorithm a glass half full of promise? Image created with Midjourney. OpenAI announced a new flagship algorithm GPT-4o that can “reason” across audio, vision, and text in real time. Like many OpenAI releases, yesterday’s announcement of
Thanks for the detailed insights, Geoff. I appreciate your reasonsed thinking about the improvements and new capabilities. I tried it yesterday and found it to be faster (even with heavy loads following the announcement). I'm wondering if I should continue paying for the Plus version, though.
🚨 Noyb files a complaint against OpenAI
All you need to know about tech policy & regulation | Luiza's Newsletter #101
2023 Annual Letter
Social Capital Performance Summary To the supporters and friends of Social Capital: This is the sixth of our annual letters where we share our reflections, key observations, and learnings over the past year, including how the economic and technological trends of the year have shaped our thinking and our investment portfolio.
For Fund II, how is the DPI higher than TVPI?
Thank you for writing. We know that government always has an incentive to spend money on new technology. In the case of AI, do we think US spending will make any difference in success versus failure? We know if they regulate AI, it will hurt.
What happened in Marketing: AI Ads for X & Meta, Reddit Search + Unusual partnerships
This Week: AI marries ad creative, Retail media hooks up with new partners and marketers blame Temu.
Happy Mother’s Day to everyone. If you are not a mother, give my wishes to your mother. Be Kind because marketing isn’t. Too many updates…. Hi, writing the newsletter takes time and efforts. If you do like to support my work. You can by choosing the option below to join the paid newsletter. (It’s your decision, I’m not your boss).
Lots happening in search and AI. I wonder if SEOs need to start thinking more in terms of being “found” on multiple platforms instead of optimising for search engines. Consumer behaviour and tech are colliding to broaden the search and consideration step in the funnel.
Suuuuuper sad about Sparks & Honey-- incredible group of people truly dedicated to understanding the signals.
LWiAI Podcast #165 - Sora challenger, Astribot's S1, Med-Gemini, Refusal in LLMs
China unveils Sora challenger able to produce videos from text similar to OpenAI tool, Capabilities of Gemini Models in Medicine, and more!
Our 165th episode with a summary and discussion of last week's big AI news! Email us your questions and feedback at contact@lastweekin.ai and/or hello@gladstone.ai Subscribe Apple Podcasts Spotify YouTube RSS Timestamps + links: Tools & Apps (00:01:27)
There's a fierce and interesting review of Github Copilot Workspace there:
The link to "Better & Faster Large Language Models via Multi-token Prediction" should be:
GC, a16z Capture 44% VC fundraising💰, Massive Acquisitions in Software Startups 🛒, Network Effects🕸️
Welcome to The VC Corner, your weekly dose of Venture Capital and Startups to keep you up and running! 🚀 You can now become a premium subscriber and read the full guest posts I share on The VC Corner. Next Saturday, I will have Peter Walker, head of insights at Carta, publishing in my newsletter a deep dive into actual VC Valuations
So valuable, as always!
Related OpenAI Substacks
