Nicolas Boitout, PhD on LinkedIn: Meta and Elon Musk’s xAI fight to partner with chatbot group Character.ai (2024)

Sam Altman twitted: "Her" during the day of the release of #GPT-4o last week.It is undeniable that "Sky" - the female voice available with the new #AI model - is so close to Scarlett Johansson's AI voice in the movie "Her" that we could legitimately call it a #deepfake. Sky has been removed from the list of the 5 available voices. ---The Daily Show did a great parody of this flirty female voice:"New ChatGPT Is Sounding Really Horny" (link in comment)

11

3 Comments

The #AI ecosystem of startups was fretting before this week's announcement of Google and OpenAI.We discovered that OpenAI has pivoted from its race to #AGI.(Sam Altman didn't seem much excited about that).Their new product (#GPT-4o) is trying to convince a broader customer base, with the "same level of IQ" (no progress towards AGI) but with new features:📢 it is free (!!), 📢 increasingly #multimodal (picture + text as inputs) 📢 and it emphasized voice interaction (GPT adopting quite a flirty voice).Many AI startups will be impacted (negatively) by this re-positioning of OpenAI.Example:#AI (generative AI) is expected to transform education.Many startups are now entering the field.But as OpenAI is pivoting, some of them are discovering they had "no moat".Full story with the slide deck.

19

Excerpt from this article:"One of Cameron’s most inspired casting decisions was hiring Arnold Schwarzenegger to play a cyborg in The Terminator. While he is open-minded about the role of AI in filmmaking, he does not believe that machines could ever do what Schwarzenegger did — give a human performance." “The machine could give you a plausible performance, but it won’t give you the quirky moment of creation that an actor gives that is particular to them and their life experience,” Cameron said. “We’re all generative AIs — we’re all vast data sets of every experience we’ve ever had. Generative AI can give you a bitmap and give you an image, but can’t give you an emotion.”James Cameron---Schwarzy to AI:“There are no shortcuts—everything is reps, reps, reps.”

10

2 Comments

Intensive Weekend: 🚀Shahar Hillel and I set up an AI voicebot dedicated to mentoring startup founders in record time. 🚀Our adventure began last Thursday while chatting with Alex M Dascalu, who was busy advising startup founders on that day.We asked him whether an "AI coach for startup founders" would be useful.His response? "Sure thing!"We got immediately inspired!Shahar and I (team #AI-migos) then registered for Veridion's great hackathon in Bucharest that very same day.We immediately began working on a voice bot, convinced that voice engagement would provide deeper insights and better advice to founders in the long run.Over the weekend, we focused on integrating this bot with Veridion's extensive dataset : Veridion ingests all corporate data from the web on a weekly basis. This enables us to map the competitive landscape for any startup project, based on a phone conversation.This challenge was submitted by Bogdan Iordache for the hackathon.A big shout-out to the organizers at Veridion!We named our bot "Greg", after my good friend Gregory Rondin!

37

5 Comments

In Europe, good AI models are scarce, but luckily, our VPNs are top-tier.#Llama-3

17

#Llama-3: Initial Reactions Post-Launch20240418 20:05: Scrolled through my #LinkedIn and what do I see? Yann Lecun's post about the Llama-3 launch!! The thing had already racked up 300 reactions in just 7 minutes. Talk about viral.20240418 20:50: Checked again 45 minutes later. Boom—3000 likes!➡️ Might just be the fastest spreading #AI news I've ever witnessed on this platform!I immediately dive into Meta's technical blog to see what this new AI model's got under the hood.20240418 22:30: Quick check at how Meta’s stock was doing on Nasdaq after this important announcement.And the result? Zilch. Nada. Zero movement—investors clearly snoozing through the news.Seriously? Llama is supposed to be Meta's big thing!!And let’s not forget, training #Llama-3 isn’t exactly pocket change:👯 398 ML Engineers proudly put their names on the announcement.💾 Meta’s Research SuperCluster boasts a 24,000 GPU setup.📚 Even the tiny Llama-3, 8B model, gobbled up 1.3 million GPU hours.Did some back-to-the envelope calculation:A single A100 instance is about $33/hour, each packing 8 GPUs.➡️ Just the initial training phase for the 8B model using these 24,000 GPUs for 2.3 days set them back a cool $5.3M bill.And that's just for the GPUs, never mind the large number of engineers and the cost of running a data center.----Let's zoom out and check again at the stock of Meta of the last 18 months.Meta’s stock miraculously went from ~$100 at the end of December 2022 to $500 now!!All this has been attributed to excellent financial results and ... to Mark Zuckerberg’s shift from "#Metaverse" to "open-source #AI." (=Llama).Both mega projects have a clear vision but, what about the business model?This leaves us with this question raised from the Financial Times:"How long can Wall Street’s artificial intelligence-fuelled rally continue without clear evidence that generative AI is giving a meaningful lift to business?"

19

Diffusion of #AI used to be slow. This time is different.The first time I encountered a Deep Learning system (Neural Nets) in production was in 2008.At that time, I had just joined the Proprietary Trading team at Societe Generale in Paris, where some traders were engaged in a secretive project with an external startup. The objective was to forecast financial market movements minute by minute. They had a large room full of servers located in a tower at La Défense in Paris.By the beginning of 2010, AI technologies were still immature and notoriously difficult to implement. Initially, access to reliable data was essential. Then, only hyper-profitable projects could afford the financial risks associated with these uncertain AI endeavors. Historically, only internet companies had the scale of projects necessary to experiment with these AI technologies.As a consequence, internet companies became experts in these new AI technologies.AI frameworks, such as Google's #TensorFlow, were then made open-source, broadening the diffusion of AI to new business sectors. However, you still needed to be a Machine Learning specialist to use them.Subsequently, new products were developed based on these frameworks to make them more user-friendly, including #LowCode solutions, which require limited coding.The latest #genAI wave introduces additional capabilities: direct interactions with users in natural language and some reasoning skills.It can now permeate industries within "the rest of the economy" that had remained largely untouched by AI and #DigitalTransformation:MedicineEducationLawto name a few.All our interactions with doctors, teachers, or lawyers are decentralized, resulting in inefficient data collection. Data collection remains a challenging issue as it requires trust and confidence from human users.However, as AI reaches performance levels comparable to doctors, lawyers, and teachers, it is probably only a matter of time before we see AI integrated into our daily interactions. #UbiquitousAI---Visualizing the Diffusion of AIWe can depict the spread of AI along a time x-axis. Alternatively, we can rank individual AI projects from hyper-profitable to less profitable*. Both visualizations align.The adoption of AI began where it was most immediately profitable. As AI technologies have advanced, their application has spread to the broader economy.*: This idea was inspired by Andrew Ng's conference at Stanford in the summer of 2023.

28

2 Comments

It is no secret that students (from high school to College) are using AI tools to help them with their essay.To what extend is it cheating?From my experience, grading has always been a nightmare.I have been teaching only occasionally for the last 20y. Kind of an extra work.Although I always enjoyed being in the classroom with students, grading their homework once the course was done, was a nightmare.I didn't have much time for this, and therefore I was always late sending marks back to the University.Grading is tough because it not only means to assess a student's work, but also implicitly to rank them with respect to each other.This means a lot of work. On top of that, it is well documented that examiners can have numerous biases. I gave up on this. I now prefer to assess the students by asking them to present within the classroom.Still, home essays are necessary in many cases.So, what are your thoughts? Should we explicitly tell them to stay away from tools like ChatGPT? (and hope than AI content detection tools can easily spot cheaters) Thanks for sharing your thoughts and opinions on this !

This content isn’t available here

Access this content and more in the LinkedIn app

9

20 Comments

Job of the future: Machine Learning Therapist 🤖 Timo Elliott's work is just great !https://lnkd.in/dfFiv2p5---Timo was at GoTech World in Bucharest last fall, a big event organized by Vlad-Mihai Tiucsan, EMBA and his team.We met at a great party organized by Oves Enterprise, Mihai Filip.

26

1 Comment