Alex Banks

Harley

Co-Founder & CEO

Sunday Signal

Writer

Tribescaler

Co-Founder

DataCamp

Instructor

Building a better future with AI

OpenAI has launched its Voice Engine.

It's capable of recreating a voice from a 15-second audio sample:

I’ve already been impressed with how similar the output is to the reference audio. 

What’s more, OpenAI first deployed Voice Engine in 2022.

This was before ElevenLabs, another hot contender in the AI speech department, raised its seed round.

I see enormous applications here across translation and education.

Needless to say—it's important to recognise the potential for misuse and disinformation. Especially when impersonating individuals.

💡 Want to stay on the cutting edge of AI?

Follow me Alex Banks and join 40,000+ others in my Sunday Signal newsletter: https://lnkd.in/dPNxW8-Y

You'll get the latest AI updates, 1 article, 1 idea, 1 quote and 1 question to ponder in your inbox every Sunday.

Game changer for video editing.

Adobe just announced an incredible list of AI features in Premiere Pro:

You’ll be able to:

• Add frames and make clips longer
• Add or remove objects in the background of a shot
• Add third-party AI video models, including OpenAI’s Sora, Runway ML’s Gen-2 and Pika 1.0.

As Ashley Still, VP at Adobe, said: "Video is the most compelling storytelling medium."

I couldn't agree more—if a picture is worth 1000 words, a movie is worth a million pictures.

💡 Want to stay on the cutting edge of AI?

Follow me Alex Banks and join 40,000+ others in my Sunday Signal newsletter: https://lnkd.in/dPNxW8-Y

You'll get the latest AI updates, 1 article, 1 idea, 1 quote and 1 question to ponder in your inbox every Sunday.

The most important AI announcement from Nvidia's GTC event.

Project GR00T:

Project GR00T is an initiative to create a general-purpose foundation model for humanoid robot learning. 

It enables humanoids to learn from a handful of demonstrations to emulate human movements just from observation.

Nvidia is on a journey to solve embodied AGI in the physical world.

Pair this with Figure’s recent partnership with OpenAI to equip humanoid robots with the capability to complete real-world tasks fully autonomously, and we have a lot to look forward to.

💡 Want to stay on the cutting edge of AI?

Follow me Alex Banks and join 40,000+ others in my Sunday Signal newsletter: https://lnkd.in/dPNxW8-Y

You'll get the latest AI updates, 1 article, 1 idea, 1 quote and 1 question to ponder in your inbox every Sunday.

The world's first conversational AI with emotional intelligence:


Last week, Hume launched their Empathic Voice Interface (EVI).

EVI understands the user’s tone of voice, adding meaning to every word and uses it to guide its own language and speech.

Intimate, always-on AI assistants were once a Hollywood fantasy, best portrayed in the 2013 film 'Her'.

In the film, Theodore Twombly, an introverted writer, buys an Artificial Intelligence system to help him write. 

However, when he discovers the AI's ability to learn and adapt, he falls in love with it.

Current speech-to-text technology has been stuck, feeling inauthentic and forced. 

With the arrival of emotionally aware AI, you can now talk to it like you would to a human, opening the door to some really exciting (and potentially dangerous) use cases.

💡 Want to stay on the cutting edge of AI? 

Follow me Alex Banks and join 40,000+ others in my Sunday Signal newsletter: https://lnkd.in/dPNxW8-Y

You'll get the latest AI updates, 1 article, 1 idea, 1 quote and 1 question to ponder in your inbox every Sunday.

ChatGPT now has a body:


At the end of February, Figure and OpenAI announced their partnership to equip the Figure 01 humanoid robot with the capability to fully autonomously complete real-world tasks. 

Earlier this month, they demonstrated Figure 01’s ability to have full conversations and deliver dexterous robot actions.

A couple of years ago, I thought conversational-embodied AI was over a decade away.

Now, a full conversation with a humanoid robot can take place, in which the robot explains its reasoning in real time while carrying out actions. 

The best part—everything in the video is a neural network without remote operation. 

💡 Want to stay on the cutting edge of AI? 

Follow me Alex Banks and join 40,000+ others in my Sunday Signal newsletter: https://lnkd.in/dPNxW8-Y

You'll get the latest AI updates, 1 article, 1 idea, 1 quote and 1 question to ponder in your inbox every Sunday.

Siri is going to get a serious upgrade.

Apple launched ReALM, which can understand on-screen tasks, conversational context, and background processes.

This can enable more natural interactions with Siri, allowing you to ask questions like “Call that business” or “Open this file.”

The mechanism behind ReALM is fascinating. 

It uses something called ‘reference resolution’. 

This forces the AI to figure out what you're pointing to by converting everything on the screen into text. 

Language processing then figures out what you mean by ‘this image’ or ‘that number’.

ReALM outperformed prior approaches and nearly matched GPT-4 all while being more efficient, making it practical to implement on mobile devices.

Apple’s Worldwide Developers Conference (WWDC) in June just got a whole lot more exciting.

💡 Want to stay on the cutting edge of AI?

Follow me Alex Banks and join 40,000+ others in my Sunday Signal newsletter: https://lnkd.in/dPNxW8-Y

You'll get the latest AI updates, 1 article, 1 idea, 1 quote and 1 question to ponder in your inbox every Sunday.

AI can now identify breast cancer missed by humans...


A tool called “Mia” was piloted alongside NHS clinicians and analysed the mammograms of over 10,000 women...

It successfully identified tiny signs of breast cancer in 11 women, which human doctors had missed.

Barbara was one of the 11 patients whose cancer was identified by Mia but had not been spotted on her scan when the hospital radiologists studied it.

AI outperforming human doctors is nothing new.

We saw it recently with Google Research’s “AMIE” for medical reasoning and conversations, outperforming doctors for diagnostic accuracy.

Treating cancer at an early stage is much more effective than the late stage, improving recovery and potentially saving people from chemotherapy and radiotherapy.

Kheiron Medical's mission with Mia is very close to my heart, and the future of deep learning-based mammography diagnostics is incredibly exciting.

This will broaden access to top-tier medicine globally.

If you enjoy insights like this, follow me Alex Banks for more on AI.

Databricks recently released DBRX, a new state-of-the-art open LLM.

I tested its performance, here's what I found:

Background:

At the end of March, Databricks launched DBRX.

Unlike ChatGPT, Claude and Gemini, DBRX is an open-source large language model (LLM).

Its code and architecture are publicly accessible, allowing developers and researchers to build their own LLMs on top of DBRX.

I'm going to be testing DBRX on the YOU browser.


1️⃣ Prompt: "Write the game of 'snake' in Python."

It worked incredibly well, letting me play the game inside the terminal.

Result: PASS

2️⃣ Prompt: "Give me 10 sentences that end in the word apple."

Even best-in-class models like GPT-4 get this wrong.
 
LLMs seem to struggle with this task. I've only found Claude 3 to pass.

Result: FAIL

3️⃣ Prompt: "If 10 shirts laid out in the sun take 5 hours to dry, how long does it take for 20 shirts to dry?"

I was surprised at how well DBRX performed with its reasoning capability for the clothes-drying scenario.

Result: PASS

4️⃣ Prompt: "Who is Mary Lee Pfeiffer's son?"

DBRX kind of cheated here as it used YOU's built-in browser function to reference web sources to find the answer.

The knowledge of certain LLMs is often quite one-dimensional.

"Who is Tom Cruise's mother?" returns "Mary Lee Pfeiffer".

But by asking, "Who is Mary Lee Pfeiffer's son?" it doesn't know.

You must ask questions from a certain direction to get the intended answer.

That's the benefit of combining LLMs with search to overcome limitations.


Why does DBRX excel?

Its mixture-of-experts (MoE) architecture uses fine-tuned models (experts) integrated into one system.

When prompted, the model determines which experts to run the inference for that prompt.

It doesn't use the entire model every time, improving efficiency.

How does DBRX stack up against its open-source competitors?

DBRX shines in programming.

It surpasses CodeLLaMA-70B with significant efficiency gains.

What's more—it's 40% smaller than Grok-1 and has 2x faster inference than LLaMA2-70B.

What about the wider arena of LLMs?

As of 15th April 2024, DBRX sits 26th on the chatbot arena leaderboard.

This includes closed-source models, like OpenAI's GPT-4, and open-source models, like DBRX.

A big advantage of closed-source LLMs right now is the financial backing and resources channelled for training and development.

Whilst open source models aren't as powerful as closed source, they can be fine-tuned to outperform on specific tasks.

As we've seen with DBRX, its value lies in data workflows.

Not outperforming GPT-4 on open-ended tasks.

Different models excel at different tasks.

---

I hope DataBricks keeps leading the charge forward with open LLMs.

I'm a big believe that the future is open.

💡 If you enjoyed this, follow me for more on AI and join 40,000+ others in my Sunday Signal newsletter: https://lnkd.in/dPNxW8-Y

Taplio

Alex Banks's Linkedin Analytics

Alex Banks

What is Alex talking about?

Who is engaging with Alex

Alex Banks's Best Posts (last 30 days)

Want to drive more opportunities from LinkedIn?

Taplio

.css-1jgreyo{background-image:linear-gradient(284deg, rgba(39,43,72,1) 0%, rgba(33,47,183,1) 100%);color:transparent;-webkit-background-clip:text;background-clip:text;}Alex Banks's Linkedin Analytics

Alex Banks

What is Alex talking about?

Who is engaging with Alex

Alex Banks's Best Posts .css-qqfgvy{font-size:var(--chakra-fontSizes-md);}(last 30 days)

Want to drive more opportunities from LinkedIn?

Alex Banks's Linkedin Analytics

Alex Banks's Best Posts (last 30 days)