Artificial intelligence is rapidly changing many aspects of how we work and live. (How many stories did you read last week about self-driving cars and job-stealing robots? Perhaps your holiday shopping involved some AI algorithms, as well.) But despite the constant flow of news, many misconceptions about AI remain.
AI doesn’t think in our sense of the word at all, Scriffignano explains. “In many ways, it’s not really intelligence. It’s regressive.”
IT leaders should make deliberate choices about what AI can and can’t do on its own. “You have to pay attention to giving AI autonomy intentionally and not by accident,”
5,208 Comments
Tomi Engdahl says:
Arms Race In Chip Performance
https://semiengineering.com/arms-race-in-chip-performance/
Processing speed is now a geopolitical issue, which could help solve one of the thorniest problems in computing.
An AI arms race is taking shape across continents. While this is perilous on many fronts, it could provide a massive boost for the chip technology—and help to solve a long-simmering problem in computing, as well as lots of lesser ones.
The U.S. government this week announced its AI Initiative, joining an international scramble for the fastest way to do multiply/accumulate and come up with good-enough results. Behind the geopolitcal rhetoric, all are working to process enormous amounts of data at blazing speeds, often with limited power budgets because some of these systems need to operate in the field using batteries. And now they all are competing for the most effective way to process that data in the shortest amount of time.
Tomi Engdahl says:
The Promise Of GDDR6 And 7nm
https://semiengineering.com/the-promise-of-gddr6-and-7nm/
Why this new memory is so critical for everything from AI to ADAS.
Computer graphics is just the beginning of the burgeoning markets that GDDR6 DRAM will supercharge over the near term. Industry forecasts expect GDDR6 to start ramping up this year and become the mainstream graphic memory by 2024. GDDR6 revenue is expected to be in the range of $6 to $8 billion, which foretells the increasingly vast system application usage over the next few years.
This healthy growth, fueled by the graphics market, is only the start for GDDR6. The memory bandwidth bottleneck is now affecting a broad range of high-performance applications including networking, data center, advanced driver assistance systems (ADAS), cryptocurrency mining, high-performance computing (HPC), machine learning and artificial intelligence (AI).
GDDR6 DRAM, at speeds of 16Gbps (per pin) or 512Gbps of total bandwidth per DRAM device, offers these applications a memory subsystem that is in many cases 5X faster than traditional memory solutions.
Tomi Engdahl says:
Tekoäly tuli liikeanturille
http://www.etn.fi/index.php/13-news/9087-tekoaly-tuli-liikeanturille
STMicroelectronics sanoo nyt lisänneensä koneoppimisprosessorin liikeanturiin. Tämän ansiosta piiri tunnistaa liikkeitä paremmin ja oppii seuraamaan aktiivisuutta entistä pienemmällä virrankulutuksella.
Tomi Engdahl says:
Can Trump’s New Initiative Make American AI Great Again?
https://www.designnews.com/electronics-test/can-trumps-new-initiative-make-american-ai-great-again/162539007160253?ADTRK=UBM&elq_mid=7469&elq_cid=876648
President Trump has signed an executive order aimed at accelerating America’s lead in artificial intelligence.
Tomi Engdahl says:
10. Artificial intelligence (AI) has caught on in IoT in the past two years
AI has been around in some form since the 1960s but often has generated more hype than results. While some hype remains, real use cases with valuable results are emerging, particularly around machine learning (ML), as adoption steadily increases. According to our research, AI and ML are being used in 60 percent of IoT activities. What changed? Three major things have spurred the increase in the use of AI: the convergence of algorithmic advances, data proliferation, and tremendous increases in power and storage capabilities at a lower cost. For AI and ML to scale, production-grade data platforms are needed. Clearly, business leaders expect that to happen, with adoption of AI and ML expected to outpace other technologies
Source: https://www.mckinsey.com/business-functions/digital-mckinsey/our-insights/ten-trends-shaping-the-internet-of-things-business-landscape
Tomi Engdahl says:
ARMv8.1-M Adds Machine Learning to Microcontrollers
https://www.electronicdesign.com/industrial-automation/armv81-m-adds-machine-learning-microcontrollers
The latest microcontroller architecture definition from Arm—ARMv8.1-M—stirs machine-learning hardware acceleration into the mix.
Tomi Engdahl says:
Shit just got real. Generate faces with one click.
https://thispersondoesnotexist.com
Tomi Engdahl says:
New AI fake text generator may be too dangerous to release, say creators
https://www.theguardian.com/technology/2019/feb/14/elon-musk-backed-ai-writes-convincing-news-fiction?CMP=share_btn_fb
The Elon Musk-backed nonprofit company OpenAI declines to release research publicly for fear of misuse
Tomi Engdahl says:
Magic Wand Learns Spells through Machine Learning and an IMU
https://hackaday.com/2018/12/07/magic-wand-learns-spells-through-machine-learning-and-an-imu/
Tomi Engdahl says:
AI is sending people to jail—and getting it wrong
https://www.technologyreview.com/s/612775/algorithms-criminal-justice-ai/
Using historical data to train risk assessment tools could mean that machines are copying the mistakes of the past.
Tomi Engdahl says:
Next-generation Armv8.1-M architecture: Delivering enhanced machine learning and signal processing for the smallest embedded devices
https://www.arm.com/company/news/2019/02/next-generation-armv8-1-m-architecture
Arm Helium technology is a new M-Profile Vector Extension bringing enhanced compute capabilities to the Armv8.1-M architecture
Delivering up to 15x performance uplift for machine learning and up to 5x uplift to signal processing tasks on the smallest of edge devices
For next-generation Cortex-M processors, aimed at small, embedded devices, where local decision-making is required
Tomi Engdahl says:
From the Segregated South to Bell Labs to the AI Frontier
https://spectrum.ieee.org/the-human-os/biomedical/devices/from-segregated-south-to-bell-labs-to-ai-frontier
Tomi Engdahl says:
Pelastaako tekoäly meidät järjettömyyksiltä?
https://ellunkanat.fi/artikkeli/pelastaako-tekoaly-meidat-jarjettomyyksilta/
Tekoälyyn erikoistuneen konsultointiyritys Fourkindin senior advisor Atte Jääskeläinen kirjoitti sopivan ohuen kirjan tekoälystä, jotta ihmiset ymmärtäisivät, että ei tässä nyt mistään mahdottoman vaikeasta asiasta ole kyse.
Tomi Engdahl says:
What an American artificial intelligence initiative really needs
https://tcrn.ch/2SGxpEd
At a high level, the American AI Initiative seems to be headed in the right direction. We absolutely need a holistic approach that considers all the various areas that are critical to building innovative AI solutions. This seems to be an underlying concept of the Initiative, as the executive order places priority on making data available across government agencies, allocating cloud computing resources to support AI R&D and training the workforce. Commitment to AI innovation is critical to maintaining our leadership position in technology with the increasing level of global AI competition.
If the government wants to demonstrate its support for AI, it needs to commit significant funding and investment in education to retain, attract and grow the talent necessary to support such a critical industry that has the potential to define our future and truly increase American competitiveness.
Tomi Engdahl says:
This AI Tool Is So Terrifying, Its Creators Don’t Want To Release It
https://www.iflscience.com/technology/this-ai-text-generator-is-so-terrifyingly-good-at-creating-fake-news-its-creators-dont-want-to-release-it-/
Artificially intelligent tech can be phenomenally (and hilariously) terrible at completing the tasks it sets out to accomplish.
But now, researchers at San Francisco-based OpenAI have announced the development of a text generating algorithm that is so terrifyingly, mind-bendingly brilliant at its job that they’ve decided to keep the technology under wraps, concerned about the potential ramifications it could have in terms of fake news.
Tomi Engdahl says:
Artificially Intelligent Players Invent Nonverbal “Languages” to Win Card Games
https://spectrum.ieee.org/tech-talk/computing/software/artificially-intelligent-players-invent-nonverbal-languages-to-win-card-games
Machines are becoming more collaborative, both with humans and one another. Soon, we may have self-driving cars that negotiate rights-of-way and robots to assist nurses with home care. But first, they’ll need to learn to communicate, and not just through spoken language.
Tomi Engdahl says:
New dynamic dependency framework may lead to better neural social and tech systems models
https://m.phys.org/news/2019-02-dynamic-framework-neural-social-tech.html
“This dynamic dependency framework provides a powerful tool to better understand many of the interacting complex systems which surround us,” wrote Havlin and team. “The generalization of dependent interactions from percolation to dynamical systems allows for the development of new models for neural, social and technological systems that better capture the subtle ways in which different systems can affect one another.”
Tomi Engdahl says:
Has a rampaging AI algorithm really killed thousands in Pakistan?
https://www.theguardian.com/science/the-lay-scientist/2016/feb/18/has-a-rampaging-ai-algorithm-really-killed-thousands-in-pakistan
A killer machine-learning algorithm guiding the U.S. drone program has killed thousands of innocent people according to some reports. What’s the truth?
Tomi Engdahl says:
It’s 2019 and Scientists Have Created Mind-Controlled Rat Cyborgs
https://www.vice.com/en_au/article/gyad49/its-2019-and-scientists-have-created-mind-controlled-rat-cyborgs?utm_source=dmfb
A team in China figured out a way to take control of a rat and “steer” it through a maze with their thoughts.
Tomi Engdahl says:
http://www.whichfaceisreal.com
Tomi Engdahl says:
Can you guess which face is real, and which is computer generated?
https://techcrunch.com/2019/02/20/can-you-guess-which-face-is-real-and-which-is-computer-generated/
Tomi Engdahl says:
Facial Recognition Software Regularly Misgenders Trans People
https://motherboard.vice.com/en_us/article/7xnwed/facial-recognition-software-regularly-misgenders-trans-people
Human computer interfaces are almost never built with transgender people in mind, and continue to reinforce existing biases.
Facial recognition software is a billion dollar industry, with Microsoft, Apple, Amazon, and Facebook developing systems, some of which have been sold to governments and private companies. Those systems are a nightmare for various reasons—some systems have, for example, been shown to misidentify black people in criminal databases while others have been unable to see black faces at all.
The problems can be severe for transgender and nonbinary people because most facial recognition software is programmed to sort people into two groups—male or female. Because these systems aren’t designed with transgender and gender nonconforming people in mind, something as common as catching a flight can become a complicated nightmare. It’s a problem that will only get worse as the TSA moves to a full biometric system at all airports and facial recognition technology spreads.
Tomi Engdahl says:
Fool ML once, shame on you. Fool ML twice, shame on… the AI dev? If you can hoodwink one model, you may be able to trick many more
Some tips on how to avoid miscreants deceiving your code
https://www.theregister.co.uk/2019/02/21/ai_attack_transfer/
Adversarial attacks that trick one machine-learning model can potentially be used to fool other so-called artificially intelligent systems, according to a new study.
It’s hoped the research will inform and persuade AI developers to make their smart software more robust against these transferable attacks, preventing malicious images, text, or audio that hoodwinks one trained model from tricking another similar model.
Tomi Engdahl says:
Why Do Adversarial Attacks Transfer? ExplainingTransferability of Evasion and Poisoning Attacks
https://arxiv.org/pdf/1809.02861.pdf
Tomi Engdahl says:
Arm has announced it will add Helium, an M-Profile Vector Extension (MVE) to its v8.1-M architecture to make machine learning and signal processing less complicated for small, power constrained embedded systems that need to process data locally. Armv8.1-M with Helium will provide real-time control code, ML and DSP execution without compromising efficiency, says the company. The configuration removes the need for a separate digital signal processor. The Helium toolchain includes Arm Development Studio, Arm Models and software libraries such as CMSIS-DSP and CMSIS-NN.
Source: https://semiengineering.com/week-in-review-design-low-power-30/
More:
Next-generation Armv8.1-M architecture: Delivering enhanced machine learning and signal processing for the smallest embedded devices
https://www.arm.com/company/news/2019/02/next-generation-armv8-1-m-architecture
Tomi Engdahl says:
NVM ReRAM Memory Cell Targets Edge AI
https://www.eetimes.com/document.asp?doc_id=1334371
Researchers at CEA-Leti and Stanford University have demonstrated a chip that integrates multiple-bit non-volatile memory (NVM) resistive RAM (ReRAM) with silicon computing units and new memory resiliency features that provide 2.3× the capacity of existing ReRAM. Target applications include energy-efficient, smart-sensor nodes to support artificial intelligence on the internet of things or edge AI.
Tomi Engdahl says:
Using AI Data For Security
https://semiengineering.com/using-ai-data-for-security/
Pushing data processing to the edge has opened up new security risks, and lots of new opportunities.
Artificial intelligence is migrating from the cloud to IoT edge devices. Now the question is how to apply that same technology to protect data and identify abnormal activity in those devices and the systems connected to them.
This is a complex problem because AI is being used on multiple fronts in this battle, as well as for multiple purposes. The technology has advanced to the point where energy-efficient neural networks can be built on silicon, and that has raised a number of questions and issues that will need to be resolved. Among them:
What are the best approaches to keep data private or secure?
What are the best approaches to identifying and reacting to aberrations in data flow or other activity without impeding other potentially safety-critical functions?
What are the most efficient ways of adding in AI-based security without impacting overall power or performance?
The general consensus through the first half of 2018 was that AI training, as well as most inferencing, would happen primarily on massively parallel server farms. Edge devices would be collectors of raw data, but the vast majority of processing would happen in the cloud, with clean data pushed back down as needed. That perspective changed as the electronics industry began realizing just how much data would have to be moved if the data was not scrubbed, and how expensive and time-consuming that would be. And underlying all of this is concern about privacy rights for some or all of that data.
Tomi Engdahl says:
ARMv8.1-M Adds Machine Learning to Microcontrollers
The latest microcontroller architecture definition from Arm—ARMv8.1-M—stirs machine-learning hardware acceleration into the mix.
https://www.electronicdesign.com/industrial-automation/armv81-m-adds-machine-learning-microcontrollers?NL=ED-005&Issue=ED-005_20190220_ED-005_566&sfvc4enews=42&cl=article_1_b&utm_rid=CPG05000002750211&utm_campaign=23424&utm_medium=email&elq2=f3363fb86e984996bbac396824acd23c
Tomi Engdahl says:
Artificial Intelligence: You know it isn’t real, yeah?
It’s not big and it’s not clever. Well, not clever, anyway
https://www.theregister.co.uk/2019/02/22/artificial_intelligence_you_know_it_isnt_real_yeah/
Tomi Engdahl says:
The Keyword:
Google and DeepMind are using AI to predict the power output of wind farms 36 hours ahead of actual generation, boosting economic value of wind energy by ~20%
Machine learning can boost the value of wind energy
https://www.blog.google/technology/ai/machine-learning-can-boost-value-wind-energy/
Tomi Engdahl says:
Sarah Dai / South China Morning Post:
China-based AI chip designer Horizon Robotics raises a $600M round at a valuation of $3B from South Korea’s SK Group, memory chip manufacturer SK Hynix, others
AI chip unicorn Horizon Robotics raises US$600m in funding as China seeks to reduce dependence on imported semiconductors
https://www.scmp.com/tech/venture-capital/article/2187828/ai-chip-unicorn-horizon-robotics-raises-us600m-funding-china
South Korea conglomerate SK led US$600m round for Horizon Robotics in latest funding for artificial intelligence semiconductor companies
Tomi Engdahl says:
Nick Statt / The Verge:
Google says its AI-powered grammar checker, announced last July, will now be available on Google Docs for all G Suite users — The power of machine translation to help you improve your writing at work — Google today announced that its artificial intelligence-powered grammar checker …
Google expands AI-powered grammar checker in Docs to all G Suite users
https://www.theverge.com/2019/2/26/18241496/google-docs-artificial-intelligence-ai-grammar-checker-g-suite-expansion
The power of machine translation to help you improve your writing at work
Tomi Engdahl says:
Christina Farr / CNBC:
Google launches program in India to screen diabetics for eye conditions that can cause blindness, using its machine learning image analysis tools
Google launches India program to screen diabetics for eye conditions that can cause blindness
https://www.cnbc.com/2019/02/25/google-verily-launch-diabetic-eye-condition-screening-tech-in-india.html
Google and Verily are developing technologies to screen for diabetic retinopathy and diabetic macular edema.
These conditions can cause severe vision loss but are preventable if caught early enough.
There’s a big shortage of eye doctors to screen populations with diabetes, especially in countries such as India.
Tomi Engdahl says:
Teardown: Intel’s Compute Stick
https://www.edn.com/design/consumer/4461617/Teardown–Intel-s-Compute-Stick?utm_source=Aspencore&utm_medium=EDN&utm_campaign
Tomi Engdahl says:
The New BeagleBone AI
https://blog.hackster.io/the-new-beaglebone-ai-b3eea55e09f2
new board from the BeagleBoard.org Foundation, the BeagleBone AI, unveiled at Embedded World, which opened today in Nuremberg, Germany
Tomi Engdahl says:
The History of Chess AI
https://blog.paessler.com/the-history-of-chess-ai?utm_source=facebook&utm_medium=cpc&utm_campaign=Burda-Blog-Global&utm_content=chessai&hsa_src=fb&hsa_acc=2004489912909367&hsa_cam=23843424879390129&hsa_grp=23843424879380129&hsa_ad=23843424890580129&hsa_net=facebook&hsa_ver=3
Tomi Engdahl says:
The military wants to build lethal tanks with AI
https://www.zdnet.com/article/the-military-wants-to-build-lethal-ai-tanks/
The ATLAS project will give combat vehicles autonomous targeting capabilities.
Tomi Engdahl says:
Optimizing Deep-Learning Inference For Embedded Devices
https://semiengineering.com/optimizing-deep-learning-inference-for-embedded-devices/
Deep artificial neural networks (ANNs) have emerged as universal feature extractors in various tasks as they approach (and in many cases surpass) human-level performance. They have become fundamental building blocks of almost every modern artificially intelligent (AI) application, from online shop recommendations to self-driving cars.
Tomi Engdahl says:
Neural Network Teaches Drones to Navigate Cities Autonomously
https://innovate.ieee.org/innovation-spotlight/neural-network-autonomy-drone-navigation/#utm_source=Facebook&utm_medium=social&utm_campaign=Innovation&utm_content=drone?LT=CMH_WB_2019_LM_XIS_Paid_Social
Tomi Engdahl says:
OpenAI built a text generator so good, it’s considered too dangerous to release
https://techcrunch.com/2019/02/17/openai-text-generator-dangerous/
Tomi Engdahl says:
Lucas Matney / TechCrunch:
Presto, which provides restaurants with food ordering hardware and an AI platform to help them make decisions about ordering stock and staffing, raises $30M
Presto raises $30M to bring its AI platform and tabletop ordering hardware to restaurant chains
https://techcrunch.com/2019/02/27/presto-raises-30m-to-bring-its-ai-platform-and-tabletop-ordering-hardware-to-restaurant-chains/
Presto is working with restaurants to update the 21st century dine-in experience, letting customers order and pay from their table with a tablet device while also providing hardware like wearables for servers so they can be alerted when they are needed by customers.
Tomi Engdahl says:
Animal-AI Olympics Will Test AI on Intelligence Tasks Designed for Crows and Chimps
https://spectrum.ieee.org/tech-talk/robotics/artificial-intelligence/animal-ai-olympics-tests-ai-on-intelligence-tasks-for-crows-and-chimps
Tomi Engdahl says:
Introducing BodyPix: Real-time Person Segmentation in the Browser with TensorFlow.js
https://medium.com/tensorflow/introducing-bodypix-real-time-person-segmentation-in-the-browser-with-tensorflow-js-f1948126c2a0
Tomi Engdahl says:
Can Machine Learning Teach Us Anything?
https://spectrum.ieee.org/robotics/artificial-intelligence/can-machine-learning-teach-us-anything
Tomi Engdahl says:
9 Demos from ISSCC 2019
Reinforcement learning on an inference budget
https://www.eetimes.com/document.asp?doc_id=1334396
Tomi Engdahl says:
Can Debug Be Tamed?
Can machine learning bring debug back under control?
https://semiengineering.com/bigger-debug-challenges-ahead/
Debug consumes more time than any other aspect of the chip design and verification process, and it adds uncertainty and risk to semiconductor development because there are always lingering questions about whether enough bugs were caught in the allotted amount of time.
Recent figures suggest that the problem is getting worse, too, as complexity and demand for reliability continue to rise. The big question now is whether new tool developments and different approaches can stem the undesirable trajectory of debug cost.
Tomi Engdahl says:
New Design Approaches At 7/5nm
https://semiengineering.com/new-design-approaches-at-7-5nm/
Smaller features and AI are creating system-level issues, but traditional ways of solving these problems don’t always work.
The race to build chips with a multitude of different processing elements and memories is making it more difficult to design, verify and test these devices, particularly when AI and leading-edge manufacturing processes are involved.
There are two fundamental problems. First, there are much tighter tolerances for all of the components in those designs due to proximity effects. Second, as a result of those tighter tolerances, better characterization data is required. However, the behavior of these chips or their component parts isn’t always precise. This is especially true for AI chips, as well as chips that include AI, where predictability and measurability are relative terms. Two devices may start out identical and diverge over time due to different use cases or environmental conditions.
“With AI, it’s not just silicon—it’s also software,” said Raik Brinkmann, CEO of OneSpin Solutions. “You can measure what’s going on with the training set, but you have limited visibility into how the network will interpolate from that. So two chips may still perform in a similar way, but they might not be at the same accuracy. The big question here is whether it is accurate enough.”
Tomi Engdahl says:
Intel Labs Director Talks Quantum, Probabilistic, and Neuromorphic Computing
Rich Uhlig, who took over Intel Labs late last year, discusses Intel’s vision for the future of computing
https://spectrum.ieee.org/tech-talk/computing/hardware/intel-labs-director-talks-quantum-probabilistic-and-neuromorphic-computing
Tomi Engdahl says:
Europe’s AI start-ups often do not use AI, study finds
https://www.ft.com/content/21b19010-3e9f-11e9-b896-fe36ec32aece?shareType=nongift
Research shows two-fifths have no artificial intelligence programs in their products
Two-fifths of Europe’s artificial intelligence start-ups do not use any AI programs in their products, according to a report that highlights the hype around the technology.
Companies branded as AI businesses have historically raised larger funding rounds and secured higher valuations than other software businesses.
The term “artificial intelligence” has been applied to a variety of technologies, from simple computer programs that automate tasks to more complex neural networks
and machine learning algorithms. The widespread use of the term has made it difficult for venture capital investors to distinguish between actual and purported AI businesses.
Getting stuck in buzzwords is never a good thing. AI has become a catch-all phrase that’s often used flippantly
Simon Cook, Draper Esprit chief executive
Tomi Engdahl says:
https://www.uusiteknologia.fi/2019/03/06/shift-keskittyy-tekoalyn-jalkauttamiseen/