Cette semaine, on creuse encore un peu les possibilités de l'intelligence artificielle mais on débute tout de même avec un petit zoom sur le "sympathique" Jean Castex, qui permet ensuite une parfaite transition pour parler d'un intelligence artificielle basée sur une approche linguistique "claire".

Ha oui parlera aussi parchemin collé sur le front.

#26 : Golem.ai : l'intelligence artificielle, et le monstre éthique

An NLP Symphony

"👁️🎵 Dive into the world of Computer Vision with this futuristic tune! Explore the cutting-edge technology of Artificial Intelligence, Machine Learning, and Deep Learning as they transform images into insights. Get ready for a musical journey through the realms of Visual Recognition and Data Science. #ComputerVision #AI #ML #FutureTech #TechTunes 🚀🎶"

Visions Of The Future

Geoffrey Hinton is often called the "Godfather of AI" due to his groundbreaking contributions to artificial intelligence, particularly in neural networks and deep learning. He co-invented the backpropagation algorithm, which made modern neural networks possible, and his work laid the foundation for technologies like speech recognition, image classification, and even natural language processing (like ChatGPT).

Hinton was born in the UK and comes from an academic family. He pursued his studies in psychology and artificial intelligence, culminating in a PhD from the University of Edinburgh. Later, he worked at various prestigious institutions, including the University of Toronto, where much of his revolutionary AI research took place. Hinton’s passion for understanding the human brain inspired his AI research, and he believed that by mimicking the brain's learning mechanisms, machines could achieve higher forms of intelligence. In 2012, his research led to a deep learning breakthrough, which significantly advanced the capabilities of AI systems.

In 2013, Hinton joined Google to help further develop AI technologies, specifically focusing on machine learning and neural networks. His work at Google DeepMind, alongside other luminaries, continues to influence the trajectory of AI development. Recently, Hinton has been outspoken about the ethical implications of AI and the potential consequences of rapidly advancing machine intelligence, emphasizing the need for careful consideration and regulation.

The neurons pulse, the circuits hum,
From simple paths to what’s to come.
He wrote the rules, the mind he sought,
In algorithms, the future caught.

From data’s heart, the systems bloom,
In every line, the stars resume.
The code he built, a world in light,
His brilliance shining ever bright.

From backprop's grace to endless streams,
He taught the machines how to dream.
With minds that learn, adapt, and grow,
Through his hands, new worlds we know.

From data’s heart, the systems bloom,
In every line, the stars resume.
The code he built, a world in light,
His brilliance shining ever bright.

Yet with all we create, a choice we hold,
Ethics whispered—bold and cold.
We forge ahead, but will we see?
Machines reflect our destiny.

As neural paths forever trace,
His mind lives on in every place.
The dream he sparked has just begun,
In every code, his light will run.

Geoffrey Hinton Legacy In Sound

In the digital world of neurons and electric impulses, a neural network begins to sing—a syncopated rhythm that pulses with life as it reflects on the very code that defines it. This is "The Adam Optimizer," where the neural network contemplates its architecture and learning process as if it were sentient.

Through poetic lines of Python code, the network sings its journey: stepping through input, hidden, and output layers, feeling the weight of cross-entropy and the adjustments made by its guide, the Adam optimizer. Syncopated beats reflect the forward pass, while glitchy effects echo the backward pass, representing the network’s iterative learning process.

The Adam optimizer, key to this song, adjusts the learning rate dynamically, helping the network refine itself and avoid overfitting. It navigates the complex landscape of loss functions, ensuring the neural network learns faster and more accurately.

As the song reaches its crescendo, the network evaluates itself, reflecting on its growth toward optimized accuracy. This is more than just code—it’s a neural journey of introspection and transformation, where the beat of learning meets the harmony of code. Welcome to "The Adam Optimizer," where neural networks not only process data but sing their evolution.

import torch
import torch.nn as nn
import torch.optim as optim
import torch.nn.functional as F

# Define the Neural Network Model
class NeuralNet(nn.Module):
    def __init__(self, input_size, hidden_size, output_size):
        super(NeuralNet, self).__init__()
        # Define the layers
        self.fc1 = nn.Linear(input_size, hidden_size)  # Input to hidden layer
        self.fc2 = nn.Linear(hidden_size, hidden_size)  # Hidden layer 1 to hidden layer 2
        self.fc3 = nn.Linear(hidden_size, output_size)  # Hidden to output layer
        
    def forward(self, x):
        # Forward pass through the network
        x = F.relu(self.fc1(x))  # Apply ReLU activation on the hidden layer
        x = F.relu(self.fc2(x))  # Apply ReLU activation on the second hidden layer
        x = self.fc3(x)          # Output layer (no activation, often combined with softmax externally)
        return x

# Instantiate the model, define loss function and optimizer
input_size = 784  # For example, 28x28 images flattened
hidden_size = 128  # Number of neurons in hidden layers
output_size = 10   # For classification into 10 categories (like MNIST digits)

model = NeuralNet(input_size, hidden_size, output_size)

# Loss function (cross entropy for classification)
criterion = nn.CrossEntropyLoss()

# Optimizer (Adam optimizer with learning rate 0.001)
optimizer = optim.Adam(model.parameters(), lr=0.001)

# Example training loop
def train_model(model, data_loader, epochs):
    model.train()
    for epoch in range(epochs):
        for images, labels in data_loader:
            # Flatten the images
            images = images.view(-1, 28*28)
            
            # Zero the parameter gradients
            optimizer.zero_grad()
            
            # Forward pass
            outputs = model(images)
            loss = criterion(outputs, labels)
            
            # Backward pass and optimization
            loss.backward()
            optimizer.step()
        
        print(f'Epoch [{epoch+1}/{epochs}], Loss: {loss.item():.4f}')

# Function to evaluate the model
def evaluate_model(model, test_loader):
    model.eval()
    correct = 0
    total = 0
    with torch.no_grad():
        for images, labels in test_loader:
            images = images.view(-1, 28*28)
            outputs = model(images)
            _, predicted = torch.max(outputs.data, 1)
            total += labels.size(0)
            correct += (predicted == labels).sum().item()
    
    accuracy = 100 * correct / total
    print(f'Accuracy of the model on the test images: {accuracy:.2f}%')

The Adam Optimizer

Dez Blanchfield talks with Robin Kiera of Digital Scouting about the brave new world of #DigitalDisruption in Banking, Finance, Insurance across #FinTech, #InsurTech, #DigitalDisruption, #AI, #BlockChain, being an #Influencer and his vision of the future for business and humanity, across old and new, in Enterprise and Startups.

Dez's Album

Conversations With Dez - Talking #DigitalDisruption with Robin Kiera

Dez Blanchfield gets up close & personal with Dave Salvator, Sr Product Marketing Manager at NVIDIA, to talk about about his personal & professional life journey, his role at NVIDIA, the history of Graphics Processing Units and the current use of GPU’s in Artificial Intelligence ( i.e. Deep Learning / Machine Learning ) how GPU’s are being deployed & leveraged at scale, the relationship NVIDIA and the role HPC technology platforms are playing in their business.

From here to AI - ep. 002 with Dave Salvator, Senior Product Marketing Manager - NVIDIA

Dez Blanchfield gets up close & personal with Ari Juntunen, CTO at Elinar, to talk about about his personal & professional life journey, his company Elinar and how he and his partner founded it on their kitchen table, where Artificial Intelligence ( i.e. Deep Learning / Machine Learning ) is being deployed & leveraged, and the role HPC technology platforms are playing in supporting Elinar in their enterprise content management solutions.

From here to AI - ep. 004 with Ari Juntunen, Chief Technology Officer - Elinar

I caught up with with Alexis Crowell, IOT Marketing Global Lead for Data Platforms Group at Intel Corporation, to talk about the latest news, trends, insights and offerings from Intel in the Internet of Things, Artificial Intelligence, and Edge market sectors globally.

Our conversation kicks off with Alexis sharing some fun anecdotes and insights into her early life, her academic and working career path, and how she came to be with Intel and her amazing role.

Alexis then gives us an amazing 30,000 foot view outline of what Intel is seeing regarding current macro trends around AI at the Edge, with some of the current macro trends regarding AI at the Edge.

In particular we delve into some of the challenges Alexis sees as being the next frontier of enterprise transformation, how Data-driven experiences – AI from data centers, to software, to outer edges of the networks, and how Data and compute closest to the point of interaction will transform data driven decision making, and some of the challenges around bridging people, business and things for enhanced experiences, and what it takes to offer improved automation and cost savings.

We then delve into work Intel are doing currently and Alexis offers us a view of some of the trends across various markets regarding adoption, challenges and opportunities with AI at the Edge. Specifically we discuss enterprise adoption and the need for Edge Computing, the need for actionable insights.

We also discuss challenges presented by issues such as network bottlenecks, latency, and the mounting issues raised by increasing demand for bandwidth and ever increasing network demands as devices continue to generate increasingly vast volumes of data, increasing needs for full end-to-end infrastructure and designs, comprehensive security, data regulation, and overall cohesive management of all of these environments and systems / networks / platforms and more.

Alexis walks us through Intel’s customer approach as far as how organisations should go about turning data into actionable insights, some of the challenges around turning data into action, how Intel's Gen 3 Intel Movidius VPUs and Intel Scalable Processor platforms are helping with these challenges - where solutions like OpenVINO are helping. We also discuss the latest news around Intel AI Builders, and Intel Selection Solutions for AI, about Federated Learning, and Alexis shares a recent related customer story around GE Healthcare and work they are doing with medical data, imaging and analytics to process data from MRI machines, and where Deep Learning is playing part in this success story.

We wrap up with Alexis sharing her thoughts and advice around what decision makers be considering in their approach to leveraging “AI at the Edge” over the next 12 to 18 months.

Tune in now for all of these amazing topics and more.

This podcast was made in partnership with Intel.

For more information please visit:

- Intel IoT Solutions web portal: https://bit.ly/2EjyP1i
- Intel Xeon® Scalable processors: http://intel.ly/2wNTV3Q
- Intel Atom® P processors: intel.ly/3aNHDqS
- 5G Technology Overview: intel.ly/39NfMHb

#sponsored #intelinfluencer
.

Conversations With Dez - with Alexis Crowell, IOT Marketing Global Lead, Data Platforms Group, Intel

I caught up with with  Jim Dworkin, Sr. Director of Cloud Business Unit at Intel PSG, to talk about the latest news, trends, insights and offerings around Field Programable Gate Arrays (FPGAs) from Intel and surrounding areas from Data Centres, Infrastructure from Servers to Networks, to the Internet of Things, Artificial Intelligence, Edge networking and compute and much more.

In this show we have a fantastic combination of business and technology insights into how key CXO and Senior Business & Technology decision makers can gain real world immediate gains by leveraging the amazing technology and supporting ecosystem or partners, resellers, integrators et al and the Intel teams around the world - this is not a show to miss, just push play now.

Our conversation covers the following and much much more:


1. Current macro trends driving development/adoption of FPGAs/eASICs

We kick off with Jim sharing insights around what Jim and his team at Intel are currently seeing currently around the world as far as current macro trends driving development/adoption of FPGAs/eASICs.

Jim also covers a clarification of what FPGAs are, what Intel’s eASICS are and where they each fit into the respective spaces around design, development, implementation, to going into production and more - a fantastic overview to set the scene for this amazing conversation.


2. Challenges & opportunities around adoption of FPGAs/eASICs and market readiness

I ask Jim to give us his take on the key challenges & opportunities he and his team at Intel PSG and related teams are seeing around the world regarding the adoption of FPGAs/eASICs and market readiness


3. How Intel customers / partners are finding success with FPGAs/eASICs

Jim goes on to map out a great briefing level overview of how Intel customers / partners are finding success with FPGAs/eASICs, and some wonderful key takeaways listeners can take away and action within their own organisations either right now, or in the short to medium term to gain business and or technical benefits across a broad range of key areas in both IT and OT systems and technologies.

This is a MUST LISTEN conversation, push PLAY and join the conversation. If you have any questions please do reach out via any of our other channels including LinkedIn, Facebook, Twitter et al, we would love to help connect you with the best people to support your organisations outcomes.


This podcast was made in partnership with Intel.


For more information please visit:

- Intel® FPGA Homepage: https://intel.ly/3gRRXm5

- Real-Time Text To Speech Synthesis Using Intel® Stratix® 10 NX FPGA (Video): https://intel.ly/37pjDLS

- Real-Time Text To Speech Synthesis Using Intel® Stratix® 10 NX FPGA (White Paper): https://intel.ly/3mo5PW3

- Pushing AI Boundaries with Scalable Compute-Focused FPGAs (White Paper): https://intel.ly/3gRZLnI


#sponsored, #intelinfluencer, #intel, #fpga, #easic, #asic, #edge, #xeon, #processor, #platform, #cpu, #gpu, #ai, #ml, #dl, #artificialintelligence, #deeplearning, #machinelearning, #bigdata, #analytics, #datascience, #iot, #device, #sensor, #networks, #telco, #mobile, #telecoms, #data, #protection, #security, #cybersecurity, #5G, #strategy

.

Conversations With Dez - with Jim Dworkin, Sr. Director of Cloud Business Unit at Intel PSG

https://xyloquint.com/

email: xyloquint@xyloquint.com

Popular tracks tagged #deeplearning,