No account yet ?

Predictive AI

We can see what data is going into the intelligent system, we can see the most likely outcome (which is not always constant), but we don't know how the learning machine made the connection between the two.

The startup elodees intends to create a new AI technology, more efficient and effective, without bias, consuming little energy, able to explain its decisions, without human supervision, and in real time, therefore more reliable.

Predictive AI

Certification of systems based on predictive AI is not possible !

For applications requiring critical systems such as air transport, software must be certified, that is, its reliability is measured and guaranteed. However, predictive artificial intelligence tools, whose results vary depending on the data used to train them, do not allow this. It's hard to imagine that self-driving cars and AI in aerospace would be allowed without any certification. Predictive AI in aerospace will never replace the captain of an airliner due to its repeated biases.

The startup elodees intends to create a new AI technology, more efficient and effective, without bias, consuming little energy, able to explain its decisions, without human supervision, and in real time, therefore more reliable.

Predictive AI

The current predictive AI will never be 100% reliable because statistics generate errors. this AI is biased and discriminatory, therefore dangerous. The EU is asking AI designers to explain decisions made by their AI, but they will never be able to explain their algorithms due to the technology employed in today's predictive AI.

https://www.weforum.org/agenda/2021/01/ursula-von-der-leyen-european-commission-davos-agenda

The startup elodees intends to create a new AI technology, more efficient and effective, without bias, consuming little energy, able to explain its decisions, without human supervision, and in real time, therefore more reliable.

Predictive AI

The carbon footprint resulting from the construction of a contemporary predictive AI neural network (Deep Learning) is disastrous for the planet and can consume as much energy as 5 cars, from production to end of life.

The startup elodees intends to create a new AI technology, more efficient and effective, without bias, consuming little energy, able to explain its decisions, without human supervision, and in real time, therefore more reliable.

Predictive AI

AI must be a conscience to give meaning to sentences and therefore to texts and so that it can write 100% on its own a play that is coherent and understandable.

In this article, 10% of the play is still human intervention.

Without artificial consciousness in the image of Human consciousness, it will not be possible for an AI to be able to create a play without Human intervention.

https://www.01net.com/actualites/une-premiere-piece-de-theatre-ecrite-par-une-intelligence-artificielle-vient-d-etre-jouee-sur-scene-2036516.html

The startup elodees intends to create a new AI technology, more efficient and effective, without bias, consuming little energy, able to explain its decisions, without human supervision, and in real time, therefore more reliable.

Predictive AI

Another much less complex and much more reliable conscious AI technology is possible. I think we should not be locked in an ivory tower. Nothing is impossible and we aim to make it happen for you.

Excerpt from the post below "These issues should be seen as an incentive to explore different AI technologies capable of solving these challenges.".

https://www.villeintelligente-mag.fr/Intelligence-artificielle-les-defis-que-les-entreprises-vont-devoir-relever-prochainement_a1054.html

The startup elodees intends to create a new AI technology, more efficient and effective, without bias, consuming little energy, able to explain its decisions, without human supervision, and in real time, therefore more reliable.

Predictive AI

Let's stop talking about artificial intelligence with predictive neural networks (statistics) as they are currently designed.

Thanks to the internet, we have been able to access a great deal of data, a fuel necessary for training artificial intelligence neural networks. Take the example of images of cats that are usually annotated. These associations make it possible to train AIs, that is to say that deep learning algorithms have thus been able to be validated with these images of cats and their captions. The machines are able to recognize a cat with 98% success thanks to the analysis of an astronomical number of annotated cat photos.

It has nothing to do with intelligence because a child needs to see two cats to be able to recognize them for life even in unusual conditions (at night, in drawing, in all positions, whatever their position. race, …).

The startup elodees intends to create a new AI technology, more efficient and effective, without bias, consuming little energy, able to explain its decisions, without human supervision, and in real time, therefore more reliable.

Predictive AI

Let's stop talking about artificial intelligence with predictive neural networks (statistics) as they are currently designed.

In 2016, a machine (Deep Mind) beat the world champion of Go who is much more complex than failures. The number of combinations in the game of GO is much greater than that of chess.

To achieve this result, it was necessary to build an ultra-powerful machine, composed of 1200 CPUs, 176 GPUs and several TPUs.

This machine (Deep Mind) consumes more than 400 kWh, or approximately 20,000 times more than a human brain, which consumes only 20 Wh.

this gigantic consumption of energy was focused on a single goal, that of taking control of the game of Go.

This machine (Deep Mind) can do nothing else.

Go players, meanwhile, do other things in their lives.

Which proves that these algorithms have nothing to do with intelligence.

This is why we propose to create a new AI technology, more efficient and effective, without bias, consuming little energy, able to explain its decisions, without human supervision, and in real time, therefore more reliable.

For this, we are looking for investors who invest in the long term in the R&D of our projects.

The startup elodees intends to create a new AI technology, more efficient and effective, without bias, consuming little energy, able to explain its decisions, without human supervision, and in real time, therefore more reliable.

Predictive AI

Investigations and reports - Work on demand

AI is aided by Human Intelligence, annotating images is a job that cannot be done by AI or a computer.

To carry out this image annotation work, workers disappear behind an algorithm, the value calculation is carried out by erasing their contributions. The platforms thus make believe that the work was done as if by magic by an AI or a computer.

Irresponsible AI and low-paid odd jobs - A predictive AI dependent on platforms to recruit workers for odd jobs online image annotation to cite just this example, but it has plenty.

Amazone - Amazon M-Turk - https://www.mturk.com Elle a été la première plateforme permettant de recruter des travailleurs pour des petits boulots en ligne.

The survival of current predictive AI with its convolutional or non-convolutional neural networks and its automatic learning requiring incredibly gigantic annotated image banks is dependent on human-assisted computation, because there are plenty of cases where this contemporary predictive AI cannot make a decision on its own. but one of the aptitudes which is unique to Humans is to be able to make a decision without prior information

The startup elodees intends to create a new AI technology, more efficient and effective, without bias, consuming little energy, able to explain its decisions, without human supervision, and in real time, therefore more reliable.

Predictive AI

Project Turing : Microsoft and Nvidia create a language model that has 105 layers and 530 billion parameters and requires 280 A100 GPUs, but the system is still skewed.

https://intelligence-artificielle.developpez.com/actu/327971/Project-Turing-Microsoft-et-Nvidia-creent-un-modele-de-langage-qui-a-105-couches-et-530-milliards-de-parametres-et-necessite-280-GPU-A100-mais-le-systeme-est-encore-biaise/

Predictive AI

Today's AI is statistical AI that only uses big data.

The AI of yesterday consisted of expert systems, logic, it used much less data and was more energy efficient.

It may be necessary to return to AI that would be hybrid.

We must turn to sobriety in terms of data, the data is there in too many quantities. We will have to sort them to go to small data.

As for computing power, we must stop systematically turning to the Cloud.

Why that ? Because the question of energy is a gigantic challenge for the planet that we will soon have to take up and which will require sacrifices.

https://www.larevuedudigital.com/en-intelligence-artificielle-faites-des-mvp-et-pas-des-poc-conseille-luc-julia-de-renault/

Predictive AI