by Andrea | May 10, 2023 | Allgemein
Why not outsource expertise, data collection and development?
The cost and time required for robust AI training can vary widely depending on several factors, including the complexity of the task, the amount and quality of data available for training, and the expertise and resources of the team involved.
In general, building a robust AI model requires a significant investment of time, effort, and resources. Some estimates suggest that developing a state-of-the-art deep learning model can take months or even years of work by a team of skilled researchers and engineers. The cost of such a project can also be significant, ranging from hundreds of thousands to millions of Euros, depending on the scope and complexity of the project.
Factors that can contribute to the cost and time required for robust AI training include:
- Data collection and preparation: Gathering high-quality data for AI training can be a time-consuming and costly process, especially for complex tasks that require large and diverse datasets. Data cleaning, formatting, and preprocessing can also add significant time and cost to the project.
- Hardware and infrastructure: Training deep learning models requires significant computing resources, including powerful GPUs, memory, and storage. The cost of these resources can be substantial, and setting up and maintaining the necessary infrastructure can also require specialized expertise.
- Expertise and personnel: Building robust AI models requires a team of experts with a range of skills, including data science, machine learning, software engineering, and domain expertise.
- Iterative development and testing: Developing a robust AI model often requires an iterative process of training, testing, and refining the model. Each iteration can require significant time and resources, especially if the team needs to collect new data or make significant changes to the model architecture.
In summary, building a robust AI model can be a significant investment of time, effort, and resources, with costs ranging from hundreds of thousands to millions of euros. The exact cost and time required will depend on the specifics of the project and the expertise and resources of the team involved.
Outsource data collection and training to Spiki
It becomes clear that creating a robust neural network is both costly and time-consuming. Collecting and processing the input data needed, and training, testing and retraining the model are huge challenges for companies not specialised in this field. So why not leave those tasks to Spiki?
We have developed a unique approach to limit the amount of data needed for our neural network training and either source the data ourselves, or help you take the correct measurements and samples in a predefined and clearly specified manner. Thus we can considerably limit time and efforts needed from your side. Our clients get a fully trained neural network model, which can be deployed via microcontroller, FPGA, as a cloud service or even ASIC as a future step in our product development.
Excited? Get in touch and learn how to unlock the full potential of your business with Spiki’s AI you can trust.
by Andrea | Apr 25, 2023 | Allgemein
Huge market chances across all industries
According to a recent McKinsey Technology Trends Outlook1 on Applied AI (artificial intelligence), intelligent applications will push frontiers and set new standards in areas such as classification, prediction and control problems. Machine learning will be of integral importance across industries, the prime use cases being computer vision and natural-language processing.
McKinsey estimate the potential global annual potential at stake from AI at $10 to $15 trillion, with supply chain management and manufacturing showing equal potential value as marketing and sales ($3 to $6 trillion), followed by service operations, product management, risk assessment, corporate finance and other possible fields of application. This potential can be attributed to expected revenue increases and cost decreases resulting from the adoption of AI. There seems to be no limit to the chances reaped by artificial intelligence.
Huge chances for various industries
The industries richest in opportunities from AI are, among others:
- Information technology and electronics
- Telecommunications
- Automotive and Assembly
- Aerospace and defense
- Healthcare systems and services
- Agriculture
- Construction and building materials
- Electric power and utilities
- Pharmaceuticals and medical products.
AI models can be used in software, hardware and electronic devices (e.g. smart home appliances), for visual simulations and pattern recognition. They can combine inputs from various sensors, thus helping to operate autonomous vehicles. Huge efficiency gains can further be achieved with the automation of manufacturing or assembly processes. Using autonomous machinery, robots and computer-vision can also enhance safety procedures and thereby mitigate risks in various processes and stages of the supply chain. There are also no limits to the possibilities of customizing products and services by using data analysis, pattern recognition and predictive tools supported by AI.
Any business can profit from the targeted use of AI. So can you!
Take off with Spiki
Spiki enables you to implement a fully trained high-performing neural network, ready to use on the hardware of your choice.
Our most mature IP service is an intuitive speech command recognition for customer-specific digital navigation needs. Spiki’s command recognition can improve ease of use and user experience over various devices and applications, such as smart phones, watches, home applications, switches, industrial control systems, UAVs or other vehicles.
Spiki’s voice recognition is offline, to keep your private details private, and requires no additional hardware. It can be deployed via microcontroller, FPGA, as a cloud service or even ASIC as a future step in our product development. Our AI can be trained to be robust against disturbances, ambient noise or low quality inputs.
Boost your efficiency and explore the opportunities of a tailormade AI to enhance ease of use, customer satisfaction, speed and safety of your appliances.
Excited? Contact us!
1 McKinsey & Company, McKinsey Technology Trends Outlook 2022: Applied AI, available at https://shorturl.at/dswA5.
by Andrea | Apr 11, 2023 | Allgemein
Spiki thrives for enabling Spiking Neural Networks.
Why?
Because they are fast.
They save energy.
They save money.
They work like a human brain.
Commercially they are not yet available, which does not prevent us from researching and trying them.
The development of artificial intelligence (AI) is progressing rapidly. The next generation of AI is already on the rise. With this generation, we are bringing neural networks closer to human thinking and learning behavior. Because we are using a well-known concept from neuroscience here – neurons and the function of spikes.
A neural network is a way of machine learning mimicking the interconnected neurons of the human brain.
third generation AI is on the rise
The first generation of artificial intelligence was a rule-based imitation logic to make reasoned conclusions within a specific and limited problem. This type of application is ideal for monitoring or optimizing a process, like this: information input –> process output.
The second AI generation is called a deep learning network (DLN). DLNs can process large amounts of structured or unstructured data and can identify patterns in video, image, text or sound data. Real life applications are financial predictions, disease mapping, user / customer behavior analysis etc. With their interpretative capacity, DLNs approximate the computation efficiency of the human brain. A further reduction and miniaturization of their energy consumption for edge applications is underway.
The upcoming third generation of AI, which we describe below, takes a further leap from literal evaluations of input data to quasi understanding. This AI takes information just as a human does, as a sequence of events and not just a snapshot of reality like DLNs. This advancement of perception is based on neural network training and inference.
SNN: energy efficient learning process close to human thinking
As stated before, one of the major challenges in neuromorphic research is to mimic human flexibility. The ability to learn from unstructured stimuli while being as energy efficient as the human brain is a great challenge. The computer building blocks within neuromorphic computer systems are analogous to the logic of human neurons. A so-called Spiking Neural Network (SNN) is a new model for arranging these elements to physically reproduce the temporal component of the functionality of neurons and synapses. This enables greater energy efficiency and plasticity.
What to expect from SNNs? Challenges and perspectives
Currently, neuromorphic computing systems are still under investigation. So far, prototypes are being developed. The technology is quickly gaining momentum and large corporations such as Intel and IBM are involved in research projects.
Nevertheless, there are still some critical points that need to be considered so that the application of neuromorphic systems can be successful. On the hardware side, we are faced with limited memory bandwidth, which is the transmission speed between memory and processor. “In-memory-computing” can avoid this kind of bottleneck. As of yet, SNNs have been considered too difficult to train for commercial applications.
With Spiki’s training method we aim to enable SNN use in specific application areas and further include robustness and formal verification. With formal verification we aim to formally guarantee important properties of neural networks such as robustness, safety, and correctness. All of these are fundamental to make SNNs fit for real life and even fit for safety-critical applications.
On the software side, we must formulate use cases and define relevant problems in which neuromorphic systems can be used to generate enough training data. The problems solved by AI technologies so far are specifically formulated for use cases that don’t use neuromorphic computing. The neuromorphic approach gives us a different, promising perspective. Compared to other high-performance computing resources, neuromorphic systems offer faster speeds while consuming less power.
Surge forward, with Spiki’s robust AI tailored to your needs
Spiki wants to surge forward and join forces with companies in robotics, telecommunication, industrial automation, sensor engineering or hardware development to reap the benefits of SNNs for innovative use cases. We offer an innovative training framework and IP for supervised learning with audio, video or continuous sensor data. Our trained networks can be made robust against low-quality input data, such as white noise, slurred speech or imperfect material surfaces. We are striving to build trustworthy AI, which can be deployed via microcontroller, FPGA, as a cloud service or even ASIC as a future step in our product development. You profit from an easy to use toolchain, from training to deployment, ready for third-party hardware.
Excited? Contact us!
by Andrea | Mar 29, 2023 | Allgemein
Where does your voice assistant keep your data?
Voice assistants make our lives so much easier: simply activating our lights, dropping a phone call or playing our favourite music by just using speech commands, hands free. Several devices, assistants and routers provide us with formidable convenience at our spoken request. Of course, an active Internet connection is required for that. Sometimes different web apps come into play to further finetune and navigate tasks. It’s all quite simple and intuitive. Nothing to worry about.
Really?
Do you value, beside the mere convenience of using voice assistants, your data privacy and security? Where does your voice assistant keep your data? Does it listen to you ALL THE TIME, storing, processing or even sharing the most personal details?
Risks of online voice processing devices
Indeed, every user shares a great deal of information, even when only talking privately at home. Usually, these recorded data are stored and processed in a cloud. This poses a risk to security, safety and privacy.
- Firstly, hackers and cybercriminals can gain access to data stored in a cloud system. Sensitive information could get into the wrong hands, e.g. access codes, passwords, financial or health information etc.
- Secondly, a user’s voice is regarded as biometrical data which can identify a human being. Voice recordings stored locally are not in danger, but stored in a cloud they could get subject to abuse. Big tech corporations frequently store data generated from voice recognition devices in a cloud, analyse and use them in order to improve their software or their advertising.
- Here is a prime example: “Amazon.com Inc. must produce millions of documents in response to discovery requests in a potential class action over the marketing of its Alexa-enabled devices and their recording of users’ conversations, a federal judge ruled”, reports Christopher Brown on Bloomberg Law (see article).
- Improper or not – some big tech companies already offer their users to opt in and allow their voice recordings to be stored, while others require them to actively opt out if they do not want this. Feel free to check for yourself the terms of use of your trusted voice assistant to stay in control.
- Any company’s information is a top security priority. In the wake of the COVID pandemic, many corporate meetings had to be transferred from personal to the online world. Video conference tools mainly use end-to-end encryption, which secures privacy and security of the online communication. Still, the encrypted video and audio content is stored in the cloud…you see where this is going, check first bullet point.
In summary, data privacy and cyber security are major challenges for both private and corporate users of voice processing software and devices using speech commands.
How to guard your privacy and security
There is only one reliable solution in order to protect your sensitive information: to store and process voice data locally or at the edge, which means closer to the device used. Beside keeping data secure, this would also solve latency issues slowing down the processing of data stored in a remote cloud server.
Spiki’s voice command recognition keeps your data locally and guards your privacy.
Relying on Spiki’s robust voice command recognition, you eliminate privacy and compliance risks and keep stored data safe. Reducing latency will reduce computation time. Since our IP can run on any third-party hardware, you save time and implementation costs. Spiki’s voice recognition is offline, to keep your private details private.
- It works without Internet connectivity
- It ensures privacy of customer data
- It is usable in noisy environments, robust and reliable to fulfil your specific requirements
- It requires no additional hardware (e.g., Alexa).
Excited? Get in touch for receiving a demonstration!
by Andrea | Mar 14, 2023 | Allgemein
Reliable performance is the key
Robustness in AI can be described as predictive certainty of machine learning systems. Robust machine learning systems perform just as they have been trained to, even in unfamiliar settings, and minimise vulnerability to adversarial attacks. Put in other words: a robust AI can detect if input data is meaningfully different to what it has been trained on and mitigate unintended effects. Robustness is therefore a key prerequisite for deploying AI in safety-critical settings.1 The European Commission has, to mitigate possible negative effects of AI on society, established a set of principles for secure and trustworthy AI. Core requirements such as the concepts of explainability of AI systems and the aforementioned robusntess will also feature in future regulations of such technologies alongside the cybersecurity of digital systems and the protection of data.
When is a machine learning system robust enough for the real world?
Imagine a system for image classification that has to determine whether the pictures or bits of pictures show cats or dogs. If slightly altered pixels, shaded spots or distorted angles of the input picture lead to a completely wrong classification, this modification is called an adversarial example. If AI models make mistakes they should not, this is the exact opposite of robustness. While funny with the cats and dogs example, this cannot stand in the real world, where, for example, an autonomously driving vehicle needs to clearly distinguish street signs and obstacles.
Robustness implies that even with perturbed inputs, i.e. with any possible alteration or minuscule change to the unperturbed input, the model still classifies it correctly and does its job just like the human brain would. In practise, verification frameworks are used to test robustness in real world situtions compared to training.
The goal is clear: safety and reliability.
Robustness solves the “never enough data” problem
Whenever an AI is trained for robustness the question arises: how much learning input, i.e. data, is enough to guarantee its functioning? The answer usually is – there is never enough, the number of datapoints or inputs is infinite. That is what makes training an AI costly and time-intensive. Robustness in training makes sure that the input data points are clearly defined and specified. This reduces the amount of input data to a finite number and saves both time and money.
Spiki makes AI robust and reliable
SPIKI is building brain-inspired AI you can trust. We developed an innovative neural networks training method for supervised learning that combines
- Robust Neural Networks
- Specification based training data for reduced data collection efforts
- Built-in formal verification for explainable AI.
We are striving to build trustworthy AI, which can be deployed via microcontroller, FPGA, as a cloud service or even ASIC as a future step in our product development. You profit from an easy to use toolchain, from training to deployment, ready for third-party hardware.
Excited? Contact us!
References
1 Tim G. J. Rudner and Helen Toner, Key Concepts in AI Safety: Robustness and Adversarial Examples, Center for Security and Emerging Technology, March 2021.
2 Hamon, R., Junklewitz, H., Sanchez, I. Robustness and Explainability of Artificial Intelligence – From technical to policy solutions, Publications Office of the European Union, Luxembourg, Luxembourg, 2020
