Archiv für die Kategorie „Künstliche Intelligenz“

Zwei Minuten aus diesem Vortrag von Tristan Harris und Aza Raskin machen gerade ein bisschen die Runde im Internet:

Der Titel dieser Stunde lautet „Das KI-Dilemma“ und wurde vor vier Tagen im Rahmen des Events summit.co hochgeladen.

In dem Beitrag beschreiben die beiden Gründer vom Center for Humane Technology eine durch machine learning trainierte Traumentschlüsselung. Also eine Person bekommt einen kurzen Animationsfilm zu sehen, der Computer dechiffriert hinterher in Textform ziemlich genau, was die Person sah. Oder auch eine ebenso trainierte Software für WLAN-Geräte, die Personen durch Funkwellen in Räumen erkennen soll.

Der Vortrag ist sehenswert und sollte aktuell bei der hysterischen KI-Diskussion nicht vergessen werden. Er ist richtig und wichtig. Nicht ganz nachvollziehen kann ich, weshalb er nur 3700 Aufrufe bisher hat. Das ist ein YouTube-Video einer Konferenz zu einem IT-Thema der Stunde und schon mehr als seltsam.

(Rate me? // Beitrag bewerten?)

Loading...

What my laptop graphics card with the help of machine learning expects from Apple as an augmented reality headset:

What the world really got:

¯\_(ツ)_/¯

(Rate me? // Beitrag bewerten?)

Loading...

Algorithms are „cool“ and run anything these days: from your smartwatch to the chatbot you are asking about stock trading options online.

From time to time we find useful algorithms for usage in a specific field; if you do what I do here, measurements about your key performance indicators (KPIs) is all around your texts. So of course this cool text iq calculator (or estimator, as it’s officially called) was of utmost interest to me:

https://writingtoiq.com/

Basically what it does: you input your (preferably English written) text with at least 50 words into the text field and hit the button „Estimate IQ of text author“. The result should be pretty self-explanatory.

For the light-package travelers, here is what to make up of the results by the author of the webapp:

„As accurate as one would expect such an algorithm to be. As „proof“ I have run the algorithm on a few examples, and you can see that the results aren’t surprising (nor should they be).
* Stephen Hawking, using his last paper as the sample, achieves an IQ of 151.
* Alexander Hamilton, using Federalist paper number 6 as the written sample, achieves an IQ of 166.
* Testing different CNN reporters using their articles, I got IQs ranging from 102 to 127.
* Testing different 5G-coronavirus conspiracy theorists using posts from one of their Facebook groups, I got IQs ranging from 75 to 89.“

Let’s run it and see how I’m doing there, shall we? My last five English language contributions here:

https://gizmeo.eu/banksy-in-lisbon/ – Estimated IQ: 102 (above average)
https://gizmeo.eu/armageddon-sounds/ – Estimated IQ: 110 (above average)
https://gizmeo.eu/review-stray/ – Estimated IQ: 107 (above average)
https://gizmeo.eu/neuromancer-vs-my-gpu/ – Estimated IQ: 135 (genius) (hint: I had quotes in here from Cyberpunk mastermind himself in my text, William Gibson…)
https://gizmeo.eu/cpu-dreams/ – Estimated IQ: 101 (above average)

Not bad for a gadget weblog! And English isn’t even my primary language. Take that, CNN, harhar!

Final verdict: if you want to have a quick check if what you are reading in English is not done by a full retard, https://writingtoiq.com/ could be very useful. I didn’t find a source code reference and could do a quick check of the algorithm but the app does what it promises. I can recommend it.

(Rate me? // Beitrag bewerten?)

Loading...

Wie meine Grafikkarte sich Femdom im Weltall vorstellt – dank Stable Diffusion

Sollte es irgendwann mal darum gehen ernsthaft fremde Planeten und Sternsysteme zu besiedeln, ist die Frau die bessere Wahl für Crews im Weltall, fand die ESA heraus: sie verbrauchen weniger Kalorien. Weniger Sauerstoff. Und weniger Wasser als Männer. Außerdem setzen sie weniger Kohlendioxid und Wärme frei. Körpergröße spielt darüberhinaus keine Rolle, selbst bei identischen Körpermaßen hat die Frau den effizienteren Stoffwechsel.

Quantifizieren lässt sich das alles auch: eine Weltraummission über drei Jahre mit vier Astronautinnen spart alleine 1,7t an Nahrung. Man hätte 2,3 Kubikmeter mehr Platz. Und spart 158 Millionen US-Dollar bzw. 141,5 Millionen Euro bei der Mission. Zu hart, das!

Wer hätte das gedacht: Ellen Ripley ist doch realistischer als man(n) denkt! EZ!

(Rate me? // Beitrag bewerten?)

Loading...

Wenn du weißt, dass du als Unternehmen gerade gefickt wirst, holst die Gründer zurück:

„ChatGPT hat den Suchmaschinenriesen auf dem falschen Fuß erwischt: Google holt die Gründer zurück, um eigene KI-Projekte voranzubringen.​“ [via]

Was wir aktuell gerade bei ChatGPT an Hype erleben ist die Wiederholung des Hypes um Eliza aus den 60ern des vergangenen kaputten Jahrhunderts (wir erinnern uns: der allererste Chatbot) und auch die Phänomene des menschlichen Erstaunens sind genau die selben. Sie entstammen der menschlichen Neugier, Inkompetenz, Fantasie, der Sucht nach der ultimativen Unterhaltung und dem Verlangen des eigenen Untergangs.

Dabei haben wir dringendere akute Probleme derzeit, wie die Stellung der Doomsday Clock auf 90 Sekunden vor Mitternacht; ein desolater Zustand, Premiere überhaupt seit es diese Uhr gibt und eine traurige Einschätzung vieler Wissenschaftlerinnen und Wissenschaftler weltweit für unsere menschliche Zukunft. Ein Eliza 2.0 wird dazu wenig beigetragen haben.

Es sei wie es sei, Page und Brin tun gut daran nicht M$ zu folgen und alles in die ChatGPT-Waagschale zu legen, was sie haben. Das werden sie auch nicht tun, dafür sehe ich beide immer noch als zu visionär, zu ausgebufft, zu kompetent in Sachen IT. Es wird was Gutes haben, dass die beiden darauf antworten werden. Und wieder wird es ein Wettrennen der Zukunftstechnologie nur innerhalb der USA geben, weshalb man Europa erneut wieder nicht leiden können wird. Wir verkacken hier immer noch unsere (IT-)Zukunft, wo sind eigentlich unsere EU-Brins und -Pages, wenn man sie braucht?

(Rate me? // Beitrag bewerten?)

Loading...

There’s something romantic and beautiful if you can get it right from a cyberpunk perspective when even AAA titles fail completely in our current times. I’m glad about the gaming experience we are able to have in „Stray“! Read here in short why.

Above: launch trailer for „Stray“, available since 09/20

You play as a cat!
I can’t remember this was tried often in the past but being able to perceive the world from a cat perspective is a rather unique experience. You realise during the game that there were other thoughts put into this piece of software as in „Blade Runner“ (1997) or „Cyberpunk 2077“ (2020), for example, which is nice. In 95% of the cases this works like a charm. In the other 5% you ask yourself why you can’t proceed or the developers won’t let you proceed. Because as a cat, you would have tried it the cat way there.

The cat wears a backpack!
The funny part is, you don’t get it from the beginning of the game which is fresh: of course we all need to learn how to be a cat, there’s no time for gimmicks at first. But when you get it, you learn that it even has space for your personal drone. Which acts as your companion throughout the game, can give you hints about the world and characters, as well as be the tech or hack aspect we love to have in any cyberpunk game. There are surprises too here with this drone but these I won’t spoil. Just play it. Just this: if you’re lucky you can even customise it in game, be it the skin or you get buttons to wear for done robot tasks.

There’s a button to meow!
In the beginning you think this is another gimmick in this specific game. But while you progress throughout the game you realise this is your only way of communication (of course, you are a cat!), so you fastly start to love it. It’s one of those fine small features we love from indie games. And „Stray“ surely delivers with this. Up to the point that if you own a cat and press this button, be sure your real cat is responding in a way or the other.

The soundtrack!
Sounds and music, all original. Gamers couldn’t be more satisfied here!

The world is believable!
Of course there were real life examples for the unique city you experience in „Stray“ and its story. However, the way it is built here (for the cat) as well as the cyberpunk elements all over the place just hit the nerve. You possibly won’t believe what’s it all about until the very end, when the secrets are lifted in a very great finale. I’m 40 and I was impressed, hit back really. If you think throughout the game you already saw the saddest elements, you will be surprised. Again, I won’t spoil it. Let’s just say it sent shivers down my spine.

Of course there are some negative elements too which you could mention but these are not critical for the overall game experience: it’s a bit short (10 hours if you collect all the memories), as well as technologically a bit behind with version four of the Unreal engine. But it’s an Indie game, so I didn’t care much and you shouldn’t either. If you are interested in cyberpunk topics, you’ll love it and it’s a must-have!

9/10
.: recommended! :.

(Rate me? // Beitrag bewerten?)

Loading...

:: Introduction

I think of Neuromancer as a pretty complicated novel. Written in 1984 by William Gibson it is one of the earliest pieces of literature of cyberpunk genre, part of its Kanon today, so to speak. Anyone interested in cyberpunk will encounter it sooner or later. And anyone could read something different in it. Part of it, it’s written from the future and not only covers artificial intelligence, hacking, virtual worlds/cyberspace, drugs, ROMs; you could go on and still not grasp the vastness here. After reading, I was not really sure if even Gibson understood what he published here. But we are cool with it, as this is another factor for one’s sympathy for this 271-pager.

Today, there are websites which aggregate the most popular quotes like the one from Goodreads here. I’m pretty sure with some of the text snippets alone, you will have problems not only from semantics perspective but also from the vocabulary alone. This is innovative and futuristic but hinders a reader from truly enjoying it. Nevertheless, the cryptic groundwork was taken often as the basis for some nice artwork pieces, like the 48-page graphic novel released in 1989 by Tom de Haven and Bruce Jensen. Heck, I even think, as I own the graphic novel too and read the novel first, that you might be better off with the graphic novel and leaving the book all alone. Although I won’t recommend it, as you know, we are all cyberpunks ourselves here.

:: My proof of concept

No matter how complicated and exotic the text, I thought this would be a perfect basis for some of my further experiments with machine learning artwork generation. I thought maybe it’s better just to leave my hardware alone: should my GPU make sense of these lines now! We are 2022, with engines like Stable Diffusion and a recent laptop, this sounds like a piece of cake! So I took the most popular quotes from the novel, the ones which made sense in creating something which a reader’s mind could have difficulties with, and let it run through my (laptop) GPU. I picked the best one from three which were created simultanously per quote in approx. three minutes. Without further ado, let’s check the results here! You can click on the pictures to see the original size. Enjoy!

:: Neuromancer vs. my GPU

„Cyberspace. A consensual hallucination experienced daily by billions of legitimate operators, in every nation“

„His eyes were eggs of unstable crystal, vibrating with a frequency whose name was rain and the sound of trains, suddenly sprouting a humming forest of hair-fine glass spines“

„We have sealed ourselves away behind our money, growing inward, generating a seamless universe of self“

„Night City was like a deranged experiment in social Darwinism, designed by a bored researcher who kept one thumb permanently on the fast-forward button“

„Don’t let the little fuckers generation gap you“

„Lost, so small amid that dark, hands grown cold, body image fading down corridors of television sky“

„We monitor many frequencies. We listen always. Came a voice, out of the babel of tongues, speaking to us. It played us a mighty dub“

„And in the bloodlit dark behind his eyes, silver phosphenes boiled in from the edge of space, hypnagogic images jerking past like a film compiled of random frames“

„Case shuffled into the nearest door and watched the other passengers as he rode. A pair of predatory-looking Christian Scientists were edging toward a trio of young office techs who wore idealized holographic vaginas on their wrists, wet pink glittering under the harsh lighting“

„His vision crawled with ghost hieroglyphs, translucent lines of symbols arranging themselves against the neutral backdrop of the bunker wall. He looked at the backs of his hands, saw faint neon molecules crawling beneath the skin, ordered by the unknowable code“

„The Matrix has its roots in primitive arcade games,‘ said the voice-over, ‚in early graphics programs and military experimentation with cranial jacks.'“

„INTO HER DARKNESS, a churning synaesthesia, where her pain was the taste of old iron, scent of melon, wings of a moth brushing her cheek. She was unconscious, and he was barred from her dreams“

„Rain woke him, a slow drizzle, his feet tangled in coils of discarded fiberoptics. The arcade’s sea of sound washed over him, receded, returned. Rolling over, he sat up and held his head“

„To call up a demon you must learn its name. Men dreamed that, once, but now it is real in another way. You know that, Case. Your business is to learn the names of programs, the long formal names, names the owners seek to conceal“

„It was a vast thing, beyond knowing, a sea of information coded in spiral and pheromone, infinite intricacy that only the body, in its strong blind way, could ever read“

„Wintermute was a simple cube of white light, that very simplicity suggesting extreme complexity“

:: Outro

I was happy that my POC was a success, of course. And I was also pretty convinced by what my graphics processor could make of those rather difficult novel lines from the 80s in our current times. However, I also see some difficulties when StableDiffusion tries to render certain things: like human appearance, as oftentimes it outputs strange limbs and also faces, especially around eyes. Same goes for text, as seen in some of those (rather grotesque) examples. Both problems could be shared with other tools like DALL-E 2 or Imagen though, I never tried those before. It’s a very young brand new technology of course, this is common. From artwork perspective, I also am not sure if this is really a problem. Furthermore, I think this is a perfect example for what we do from read-write culture. Which can’t be emphasised enough nowadays, because what we originally intended back then, generate reciprocal relationship between the producer and the consumer, wasn’t taken up by people as we intended. Like ever. Stupid, unreasonable consumerism won, unfortunately. As such, this might be not the last project of its kind I did by myself, as I got enough books in my private library which could need clarification visually. But in the history of the web, probably very much so. Let me know what you think in the comments, hope you had a great time exploring this!

(Rate me? // Beitrag bewerten?)

Loading...

My laptop CPU became an artist over night:

What we see here is actual unique artwork my CPU created over a timeframe of approx. 12 minutes. The only thing I input to was the text string a cyberpunk cityscape and let Stable Diffusion do the rest. It’s beautiful and detailed. I actually love it!

Stable Diffusion is a sophisticated, brand new deep learning model by Stability.ai. It’s already on par with the earlier released, proprietary AI image generators DALL-E 2 or Imagen. With a powerful difference: you can actually create AI images yourself for free on your local machine. Usually done by your GPU, or like in my case above, your CPU, if capable.

I put four hours of work into it yesterday to setup my machine and create more artwork. My CPU came up with more beautiful cyberpunk cityscape examples with the following text string, a cyberpunk cityscape in neonlights at night, see the gallery here:

I’m lucky to own a powerful notebook CPU from 2020 with a graphics processor built in. I don’t know if this made a difference or not, the results surely are true for what I anticipated and surpassed my expectations.

But how can you join the CPU AI image generation yourself?

I had time to write it down yesterday and am happy to share it here for free for anyone who might be interested to do this with her/his Linux machine.

Step 0: Create a working directory in your /home

Step 1: Install Python

sudo apt-get update*
yes | sudo apt-get install python3.8*

*could be the installation of Python 3.8 is different to the one mentioned above. In this case just use your package manager of your Linux distribution to get the program installed on your machine.

Step 2: Install Miniconda

wget https://repo.anaconda.com/miniconda/Miniconda3-py38_4.12.0-Linux-x86_64.sh
bash Miniconda3-py38_4.12.0-Linux-x86_64.sh

Step 3: Clone the Stable Diffusion Repository

git clone https://github.com/CompVis/stable-diffusion.git
cd stable-diffusion/

Hint: Repo above only works for graphics cards >= 7GB (V)RAM! Thus…

Step 4: Download the ZIP, extract, copy and paste the contents following optimised version of Stable Diffusion in the folder above:

https://github.com/basujindal/stable-diffusion

Step 5: Create Conda Environment

conda env create -f environment.yaml
conda activate ldm

Step 6: Download Stable Diffusion Weights

curl https://www.googleapis.com/storage/v1/b/aai-blog-files/o/sd-v1-4.ckpt?alt=media > sd-v1-4.ckpt

Step 7: Content blocker deactivation (needed in case your output images are green or black only)
* find and open file pipeline_stable_diffusion.py
* below image = image.cpu().permute(0, 2, 3, 1).numpy() replace anything with this and save the file:

# run safety checker
#safety_cheker_input = self.feature_extractor(self.numpy_to_pil(image), return_tensors=“pt“).to(self.device)
#image, has_nsfw_concept = self.safety_checker(images=image, clip_input=safety_cheker_input.pixel_values)

if output_type == „pil“:
image = self.numpy_to_pil(image)

#return {„sample“: image, „nsfw_content_detected“: has_nsfw_concept}
return {„sample“: image, „nsfw_content_detected“: False}

Step 8: Generate AI images with the CPU

python optimizedSD/optimized_txt2img.py –device cpu –precision full –prompt „a cyberpunk cityscape“ –H 512 –W 512 –n_iter 1 –n_samples 1 –ddim_steps 50

That’s it! By default you find your created images in the outputs folder. You can enter your deserved text string in „“ and let your machine do the rest. Feel free to share some of your works in the comments if this tutorial helped you.

P.S.: I also had the chance to create GPU images yesterday with this method. The only difference is by removing the part –device cpu from the command line code you find in step eight. It will be much faster but to me the option to let your CPU create actual artwork was very much appreciated and may not be well known for future AI generation images through Stable Diffusion. I had some very nice results. Will share more of my work in future entries here on gizmeo.eu so stay tuned!

(Rate me? // Beitrag bewerten?)

Loading...

In der Diskussion um die derzeit in weltweiten Schlagzeilen kursierende Pseudo-KI LaMDA aus den Forschungslaboren Googles geht es um keine wirkliche Künstliche Intelligenz, weil es „wirkliche künstliche Intelligenz“ nicht gibt (siehe hierzu), zumindest aktuell. Und wahrscheinlich noch auf Jahrhunderte hinaus.

Die Software ist aber auf eine andere Art und Weise interessant, doch um die Anfänge zu verstehen muss man weit zurück gehen. Bis ins Jahr 1966 und zum alten (M)IT-Mastermind Joseph Weizenbaum nämlich.

Er nannte seinen Chatbot (auch wenn seine Software 1966 noch nicht so genannt wurde) ganz sexy und selbstbewusst ELIZA. Und ELIZA konnte bereits erste Menschen in die Irre locken, sie sei ein Bewusstsein, weil sie auf Fragen irgendwie vorgekaut reagieren konnte, was dem Bot in einer Datenbank bereitgelegt wurde. Nach dem selben Prinzip funktioniert LaMDA. Menschen waren schon immer sehr dumm und Computer extrem schwer zu verstehen. Tut mir leid.

Was wir aus LaMDA und dem flüchtenden konfusen Ex-Google-Mitarbeiter ableiten können für uns ist Folgendes:

1) ELIZA ist mittlerweile auf dem Stand eines siebenjährigen Menschen
2) es ist immer noch ein Bot und spuckt aus, was du ihm vorkaust
3) es besitzt keinerlei eigenes, gar künstliches Bewusstsein
4) es wird in Zukunft möglich sein Bots so zu programmieren, dass sie ein bestimmtes Alter vortäuschen können, da sie individuell mit entsprechenden Daten gefüttert werden können

Zu Punkt 4) noch eine persönliche Anmerkung, dass wir einen Schritt weiter zu wirklich funktionierenden Chatbots sind dank Google, die eventuell sogar mit Alter und Background programmierbar sein werden und eventuell sogar kostenlos bis wenig kosten werden.

Aber um eine „edle Version“ zu bekommen, die wie eine „wirkliche KI“ funktioniert und sogar VIP-Status(!) hat, wie bereits simuliert(!) in diversen „Deus Ex“-Videospielen gesehen (sie heißt dort, exakt in Anlehnung an Weizenbaums revolutionieren ersten Bot Eliza Cassan) werden wir entweder noch sehr lange warten müssen oder eben weiter zocken, denn mit Realität hat das leider nichts zu tun. Warum da jemand flüchten muss vor leuchtet mir aus IT-Sicht nicht ein. -.-

Oben: selbst als virtueller Charakter der Realität nach wie vor haushoch überlegen – VIP und News Anchor Eliza Cassan

(Rate me? // Beitrag bewerten?)

Loading...

This is a robotic study which is cool and sad at the same time:

„Plan Bee“ is a robotic drone industrial design study by Anna Haldewang to replace real bees by robotic ones if they go extinct so they can pollinate flowers. And extinction can go quick, unfortunately.

There! There it is again! We solve nature’s problems by technology! What can go wrong!?

(Rate me? // Beitrag bewerten?)

Loading...

Single shares of the graphics card company Nvidia are at a whopping 657€ at the moment. Read here why it’s still a very good purchase currently if you’re planning on a long-term ROI (return on investment) regarding 10-15 years in the future.

Above: a Tesla/NVIDIA GPU cluster

1) AI development: Nvidia is one of the forerunners on the development of AI, Artificial Intelligence. Not only is this an industry sector which will be transformative unlike any other for the current century (and maybe more ahead), they offer concrete hardware for development already. If you’re interested in learning it for yourself, look out for „TensorFlow“ keywords or „Tensor core hardware“ here. Equipment is available for a few hundred bucks already
2) the ARM deal: although it can be taken as a hostile takeover, the purchase of British CPU company ARM, which is currently still under investigation, will be a pusher for the company like no other before. Maybe Nvidia may be late in the game but what they lack is the CPUGPU symbiosis competitor AMD can already offer since the ATI deal back in the 2000s
3) cryptocurrency mining: obscure as this may sound, Nvidia graphics cards are among the best for cryptocurrency mining, especially popular ones like Ether. Oh, and they are a good for gaming too!
4) as long as this pandemic may last, people will play video games. And those gamers need those graphics cards, the sooner and the more popular (thus more expensive), the better

Then there are the incredible company vitals, financial aspects are the most important if you plan to invest your hard earned money into a company.

Get that: fiscal first-quarter revenue soared 84% year over year to $5.66 billion. Gaming revenue doubled to $2.76 billion. Data center revenue 79% up to $2.05 billion. Earnings per share 106% year over year to $3.03. All of the mentioned figures which was even better as prognosed by the experts in this field (lower revenue and non-GAAP earnings per share of $5.41 billion and $3.28, respectively). [via]

What the company wants to do now, as is the trend in the high-tech industry sector since the pandemic began (e.g. Apple and Tesla did this too), is to split its stock and double it in sum (from 2 billion to 4 billion), which was announced in May and approved by the share holders some weeks later. Date will be the 20th of July by a ratio 1:4; which basically means, if you hold a share in Nvidia by 19th of July, you will own 4 the day afterwards.

Companies do this for a couple of reasons, some of the most popular ones are „we are overvalued, we now help ourselves“ or „we want to enable the small investors to get a piece of us by lowering the entrance bar“. Both of it is certainly true, we shouldn’t forget though publicly listed companies want to earn more money which will certainly also happen in the case of Nvidia, albeit not in the first months or first year after split. But possibly in years and years to come afterwards. Plus, the last split Nvidia did was in 2006; as you might see easily you can become a part of history here as this will certainly not be the case for years, maybe decades, to come.

Thus fear not to get active if you can spare a few hundreds of Euros but do it quick: latest tomorrow or maybe even by July 19th which is Monday, as this is the final date the stock split goes ex-dividend. The latter translated means: sources which state that you need to hold an Nvidia share by June 21 2021 already are simply false. That is the only quirk here which was announced at Nvidia’s annual meeting of stockholders on June 3. A bit confusing, but hey: it was never easy to earn money in the real world anyway, right?

Hint: I’m not affiliated with the stock market or the company I wrote about in this article. I’m just an IT expert and want to share an important insider information about a graphics company I know since a little child. 🙂

(Rate me? // Beitrag bewerten?)

Loading...

Software.
Hardware.
Made in USA.
Nothing they create is Excellent.
Most of it is Good.
Sometimes even Very Good.
In any case, good is best.
Never try to be excellent. You will fail.
Never stop reaching for the stars.
Never stop being you.
Consume.
Adapt.
Create.
Feel good.
It sounds simple.
It’s incredibly difficult.
Just like life.
Just create. And never stop.
Never.
You’re good.

(Rate me? // Beitrag bewerten?)

Loading...

Wer mag sie nicht, unsere kreativen US-amerikanischen „Nachbarn“: „The U.S. Is One Step Closer to Establishing a Research Program to Block the Sun“, via gizmodo.

Die feiern das: das ist McDonalds gegen den Klimawandel; das in asiatischen Ländern von Kinderhänden gefertigte Nike für das CO2-Problem der Erde; die Killer-Smartphone-App, die alles kann gegen die Schmelze der Arktis. Das funktioniert ganz bestimmt!11

Ironie der Geschichte, dass selbst Hollywood, also das US-eigene Propagandaministerium für die vividen Träume der westlichen Welt, das nicht nur 2003 bereits vorwegnahm, sondern auch das Scheitern des selben. Wir präsentieren, Operation Dark Storm, animiertes Geoengineering, wenn auch zu einem anderweitigen Zweck:

Quelle: „Animatrix“, (C) WarnerBros, 2003

We solved it, now clap your skully hands, suckers!

¯\_(ツ)_/¯

(Rate me? // Beitrag bewerten?)

Loading...

Wer hochwertige Gadgets möchte, der kauft bei Samsung. Wer schon mal einen Sony Bravia-TV auseinandergebaut hat oder ein iPhone, hier sind Teile von Samsung verbaut. Chips, Displays, Inverter – you name it.

Doch auch als Innovatoren können die Südkoreaner sehr gut, ihr neustes vielversprechendes Hardware-Bauteil: ein KI-RAM.

Intern nennen sie es FIMDRAM, im Detail schaut der Die (Halbleiter-Waferstückchen) dann so aus:

Praktisch haben sie wohl „einfach“ programmierbare Recheneinheiten (Programmable Computing Units (PCU)) zwischen die Speicherbänke verbaut. Die können zwar aktuell lediglich 16-bitige Fließkommazahlen verarbeiten und verfügen nur über einfache Instruktionen (Daten verschieben; Multiplikation; Addition). Dennoch clocken diese mit 300MHz – was einer Leistung von 1,2 TFLOPS entsprechen soll. Das ist enorm: zum Vergleich, das gesamte PS4-System (keine Pro-Version) bringt 1,84 TFLOPS Leistung…

Samsungs neuartiger RAM im Detail (PPTs): https://m.gsmarena.com/samsung_creates_ram_with_integrated_ai_processing_hardware-news-47794.php.

Hardwareseitig könnte man effizientere KI-CPUs wohl aktuell nicht bauen. Die Latenzen sind hier sehr gering, ferner verbrauchen die Chips 71% weniger Energie. Ein Wermutstropfen bleibt aber noch: nicht nur wird das Design erst diese Woche bei der International Solid-State Circuits Virtual Conference von den Südkoreanern der Weltöffentlichkeit präsentiert. Die Validierung von Experten wird noch mindestens bis Juli dauern. Mit etwas Glück erscheinen diese speziellen neuen KI-RAM-Riegel dann wohl Ende dieses Jahr/Anfang 2022 in Hardware, die man als 0815-Konsument kaufen und programmieren können wird.

Ein Wendepunkt in der Programmierung künstlicher Intelligenz? Könnte sein. Ich persönlich traue es Samsung durchaus zu.

(Rate me? // Beitrag bewerten?)

Loading...

Mit einer ähnlich hohen wissenschaftlichen Relevanz wie 1996 mit dem Schachcomputer „Deep Blue“ gegen Schachweltmeister Garri Kasparow, so hat exakt 20 Jahre später ein Computerprogramm, „AlphaGo“, den südkoreanischen Großmeister des asiatischen Strategiespiels Go, Lee Sedol, geschlagen.

Eine Dokumentation in Spielfilmlänge (ca. 90 Minuten), die mittlerweile frei und kostenlos im Internet erhältlich ist, kann man sich hier anschauen:

Eigentlich galt Go bis zu diesem Zeitpunkt als von einer KI (künstlichen Intelligenz) „unknackbar“, hat es nicht nur mehr Felder pro Brett (19×19, statt 8×8), sondern lässt in Theorie mehr Züge zu, „als es Atome im Universum gibt“. Ferner ist es durch reines Herumprobieren, dem sogenannten Brute Forcing, auch nicht im Voraus berechenbar.

Es stellte sich heraus, für die Künstliche Intelligenz mussten nicht nur neue Algorithmen her (in diesem Fall der Monte-Carlo, da er mit Zufällen arbeiten kann), sondern gleich ein komplett neues neuronales Netz, das tiefe neuronale Netzwerk. Nach 9 Jahren Forschung wollte die Firma DeepMind dann genauer wissen, was die Eigenentwicklung wirklich drauf hatte: Sedol verlor vier von fünf Spielen.

Der Film ist nichts für Philanthropen, bietet allerdings neben interessanten Einsichten in menschliche Psychen auch viel Wissenswertes zum Thema Informatik und künstliche Intelligenz allgemein. Man darf gespannt sein, wo man hier 2036 angekommen sein wird; wer/welche Firma das Programm stellt; und welches Spiel die Menschheit noch für immer verloren geben muss (siehe auch: Der Alpha-Go-Schock und die Folgen, 22.12.16).

(Rate me? // Beitrag bewerten?)

Loading...