Mediaboss Marketing

This is a really amazing result, says Franois Le Gall, a mathematician at Nagoya University in Japan, who was not involved in the work. Matrix multiplication is used everywhere in engineering, he says. Anything you want to solve numerically, you typically use matrices.

Despite the calculations ubiquity, it is still not well understood. A matrix is simply a grid of numbers, representing anything you want. Multiplying two matrices together typically involves multiplying the rows of one with the columns of the other. The basic technique for solving the problem is taught in high school. Its like the ABC of computing, says Pushmeet Kohli, head of DeepMinds AI for Science team.

But things get complicated when you try to find a faster method. Nobody knows the best algorithm for solving it, says Le Gall. Its one of the biggest open problems in computer science.

This is because there are more ways to multiply two matrices together than there are atoms in the universe (10 to the power of 33, for some of the cases the researchers looked at). The number of possible actions is almost infinite, says Thomas Hubert, an engineer at DeepMind.

The trick was to turn the problem into a kind of three-dimensional board game, called TensorGame. The board represents the multiplication problem to be solved, and each move represents the next step in solving that problem. The series of moves made in a game therefore represents an algorithm.

The researchers trained a new version of AlphaZero, called AlphaTensor, to play this game. Instead of learning the best series of moves to make in Go or chess, AlphaTensor learned the best series of steps to make when multiplying matrices. It was rewarded for winning the game in as few moves as possible.

We transformed this into a game, our favorite kind of framework, says Hubert, who was one of the lead researchers on AlphaZero.

Link:
DeepMinds game-playing AI has beaten a 50-year-old record in computer science - MIT Technology Review

1 Hurricane Ian is likely to be Floridas deadliest in 87 yearsThe majority of the 100+ casualties are believed to have drowned. (WP $)+ Areas that embrace solar power fare better in extreme weather. (Slate $)+ Bangkoks flooding problem is steadily worsening. (New Yorker $)

2 Its not too late to avoid a winter of extreme illnessAccepting flu and covid shots can help to lessen the blow. (The Atlantic $)+ Covid vaccines don't harm menstrual cycles, a new study says. (Economist $)+ This nanoparticle could be the key to a universal covid vaccine. (MIT Technology Review)

3 You shouldnt worry about the US election getting hackedAt least, thats what the DBI and CISA are saying. (Motherboard)+The alt-rights tech tactics have evolved since the Capitol riots. (Slate $)+ Election misinformation is still thriving in non-English languages. (CNET)

4 Pollution particles can reach babies in the wombDepending on how much pollution the mother is exposed to, soot particles can cross the placenta. (Bloomberg $)

5 Big Tech destroys millions of data storage devices a yearEven though they could wipe and resell them, companies are scared stiff of confidential data falling into the wrong hands. (FT $)

6 Inside the race to end HIVusing CRISPRIn theory, the technology could return cells to a near-standard state. (Wired $)+ The scientist who co-created CRISPR isnt ruling out engineered babies someday. (MIT Technology Review)

7 Chinese apps are still thriving in IndiaDespite the Indian governments efforts to push users toward native apps. (Rest of World)+ Censorship-evading apps are being stamped out in China. (TechCrunch)

8 The rise and rise of facial recognition in US airportsSelf-check in kiosks are being phased out in favor of the controversial technology. (NYT $)+ If you get your face scanned the next time you fly, heres what you should know. (MIT Technology Review)

9 What its like to visit an Instagram tourist trapIt sounds like a whole lot more trouble than its worth. (Vox)

10 Its time to embrace robot dolphins Theyre an ethical alternative to the real thing in captivity. (Hakai Magazine)

Quote of the day

The spam finds its way into my inbox, too.

Commissioner Ellen L. Weintraub of the Federal Election Commission, who helps police US political campaigns, tells the Washington Post that even she cant escape the deluge of political spam emails.

The big story

The rest is here:
The Download: TikTok moral panics, and DeepMinds record-breaking AI - MIT Technology Review

Learn how your company can create applications to automate tasks and generate further efficiencies through low-code/no-code tools on November 9 at the virtual Low-Code/No-Code Summit. Register here.

This week, Googled-owned tech lab, DeepMind, unveiled its first AI that is capable of creating its own algorithms to speed up matrix multiplication. Though its taught in high school math, matrix multiplication is actually fundamental to computational tasks and remains a core operation in neural networks.

In the same vein, OpenAI this week announced the release of Whisper its open-source, deep learning model for speech recognition. The company claims the technology already shows promising results transcribing audio in several languages.

Joining the innovation sprint this week, Intel detailed a plan to make developers lives a bit easier, with a goal to make it possible to build an application once that can run on any operating system. Historically, this was a goal of the Java programming language, but even today the process is not uniform across the computing landscape something Intel hopes to change.

On the security front, enterprise leaders had several new announcements to take note of this week, including the zero-day flaw exploit in Microsofts Exchange Server. The company confirmed that a suspected state-sponsored threat actor was able to successfully exfiltrate data from fewer than 10 organizations using its staple platform.

Low-Code/No-Code Summit

Join todays leading executives at the Low-Code/No-Code Summit virtually on November 9. Register for your free pass today.

While its no secret that attacks like these continue to expand in both volume and intensity the methods for preventing attacks are also evolving. Vulnerability solutions provider Tenable is one that has evolved to change its main focus, too. This week, the company announced its shifting its focus from vulnerability management to attack surface management and released a new tool for enterprises with that focus.

Heres more from our top five tech stories of the week:

AlphaTensor, according to a DeepMind blog post, builds upon AlphaZero, an agent that has shown superhuman performance on board games like chess and Go. This new work takes the AlphaZero journey further, moving from playing games to tackling unsolved mathematical problems.

This research delves into how AI could be used to improve computer science itself.

The ability to build once and run anywhere, however, is not uniform across the computing landscape in 2022. Its a situation that Intel is looking to help change, at least when it comes to accelerated computing and the use of GPUs.

Intel is contributing heavily to the open-source SYCL specification (SYCL is pronounced like sickle) that aims to do for GPU and accelerated computing what Java did decades ago for application development.

Exposure management gives security teams a broader view of the attack surface, offering the ability to conduct attack path analysis to analyze attack paths from externally identified points to internal assets. It also allows organizations to create a centralized inventory of all IT, cloud, Active Directory and web assets.

While information is limited, Microsoft has confirmed in a blog post that these exploits have been used by a suspected state-sponsored threat actor to target fewer than 10 organizations and successfully exfiltrate data.

Developers and researchers who have experimented with Whisper are also impressed with what the model can do. However, what is perhaps equally important is what Whispers release tells us about the shifting culture in artificial intelligence (AI) research and the kind of applications we can expect in the future.

VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.

Go here to read the rest:
Top 5 stories of the week: DeepMind and OpenAI advancements, Intels plan for GPUs, Microsofts zero-day flaws - VentureBeat

A new poll finds respondents nearly twice as likely to support than oppose repealing stage three tax cuts.

The Australia Institute has a poll out which offers the interesting finding that 41% favour the repeal of the stage three tax cuts, with only 22% on board and the remainder unsure. Forty-six per cent understood the cuts to most favour high income earners, compared with 18% for middle income earners and 8% for low income earners. Asked whether adapting economic policy to suit the changing circumstances even if that means breaking an election promise rated higher than keeping an election promise regardless of how economic circumstances have changed, 61% favoured the former and 27% the latter. The poll was conducted September 6 to 9 from a sample of 1409.

The Guardian reports on the fortnightly poll from Essential Research, which continues to hold off from voting intention and does not include leadership ratings on this occasion, and is mostly devoted to questions on incidental political relevance regarding the Optus security breach. Fifty-one per cent would support stronger curbs on information collected by private companies and 47% expressed concern about governments collecting their personal information. The full report should be along later today.

UPDATE: Full Essential Research report here.

William Bowe is a Perth-based election analyst and occasional teacher of political science. His blog, The Poll Bludger, has existed in one form or another since 2004, and is one of the most heavily trafficked websites on Australian politics.View all posts by William Bowe

View post:
Taxing times (open thread) The Poll Bludger - The Poll Bludger

The AlphaGo Zero Cheat Sheet (high-res link below)

Download the AlphaGo Zero cheat sheet

Recently Google DeepMind announced AlphaGo Zero an extraordinary achievement that has shown how it is possible to train an agent to a superhuman level in the highly complex and challenging domain of Go, tabula rasa that is, from a blank slate, with no human expert play used as training data.

It thrashed the previous reincarnation 1000, using only 4TPUs instead of 48TPUs and a single neural network instead of two.

The paper that the cheat sheet is based on was published in Nature and is available here. I highly recommend you read it, as it explains in detail how deep learning and Monte Carlo Tree Search are combined to produce a powerful reinforcement learning algorithm.

Hopefully you find the AlphaGo Zero cheat sheet useful let me know if you find any typos or have questions about anything in the document.

If you would like to learn more about how our company, Applied Data Science develops innovative data science solutions for businesses, feel free to get in touch through our website or directly through LinkedIn.

and if you like this, feel free to leave a few hearty claps 🙂

Applied Data Science is a London based consultancy that implements end-to-end data science solutions for businesses, delivering measurable value. If youre looking to do more with your data, lets talk.

Read more here:
AlphaGo Zero Explained In One Diagram | by David Foster - Medium