Science journal Nature recently announced it will be launching a dedicated section of its online publication to machine learning in January, 2019. That would normally be welcome, except the section is set to be put behind a paywall.

Naturally, many AI researchers took umbrage at the plans. Much of the community are fierce advocates of open research — both to improve advancements in the field, and to help ensure the development of safe and ethical standards.

Thousands of researchers, including some of the most famous names and companies, will not submit any of their work to Nature’s machine learning section in protest. One famous AI researcher, Professor Tom Dietterich from Oregon State University, published a statement on the university’s website condemning the move.

In his statement, Dietterich highlights a mass resignation letter back in 2001 from the editorial board of the Machine Learning Journal. The editors resigned en masse to form a new zero-cost open access journal, the Journal of Machine Learning Research (JMLR).

Quoting from the 2001 resignation letter:

“…journals should principally serve the needs of the intellectual community, in particular by providing the immediate and universal access to journal articles that modern technology supports, and doing so at a cost that excludes no one.”

Furthermore, Dietterich notes most of the major machine learning outlets — including NIPS, ICML, ICLR, COLT, UAI, and AISTATS — do not charge for access to, or the publication of, AI papers.

As of writing, the petition against the move has 2,366 signatories. This includes some of the most recognisable names in AI research such as Ian Goodfellow of Google and Yann LeCun of Facebook.

You can add your signature here if you also believe AI research should be open.

Do you agree with the AI researchers?

The post Thousands of AI researchers are boycotting a science journal appeared first on AI News.

Huawei commissioned the research and found the average person is unaware of 99.68 percent of their decisions. Approximately 35,000 decisions are made per day by the human brain although, on average, respondents believe they only make around 111.

“The research shows how human intelligence works just like Artificial Intelligence, operating in the background to empower us in everything we do,” comments Walter Ji, President, Huawei Western Europe Consumer Business Group. “While revealing a significant gap between the number of decisions we believe we make every day and the actual number we make, the results also shed light on other discrepancies between how we think we spend our time, and how we actually spend it.”

Artificial intelligence can significantly improve efficiency and cut mundane tasks. That much we know, but it can also work in the background to make decisions similar to our subconscious.

The research was conducted to promote the Huawei Mate 10 Pro and its world-first AI processor — which we’re impressed with. In an opinion article, we predict an invasion of dedicated AI chips in smartphones in 2018.

Some interesting findings in the research include:

• The average guess at the number of decisions made about food each day was five, when actually it’s 221.
• The guess at the number of times we check our phones each day was 20, but in reality it’s 76.
• When making a cup of tea we think we make just three decisions, but we actually make an incredible 65.

Paul Dolan, Professor of Behavioural Science at the London School of Economics and Political Science, comments: “The results are fascinating and absolutely demonstrate that the automatic mind processes many things unconsciously. Contrary to what most of us believe, human decision-making is a process handled to a large extent by unconscious mental activity – and in ways that are actually really helpful. We create habits to prevent our minds from being overloaded by simple routine tasks.”

Most of us believe when getting ready to go out we take around seven minutes to decide what to wear. In reality, it’s around 17 minutes. Amazon, with the Echo Look, is attempting to help out with this very specific decision.

If AI can help us shave time off making some of the average 35,000 decisions per day, then more focus is able to be placed on more important tasks which cannot yet be automated.

The research was carried out in November 2017 by Lightspeed Research and features respondents from ten countries: UK, Italy, Germany, Portugal, Spain, Netherlands, Ireland, Switzerland, Belgium and France. It is nationally representative by age, gender and region.

What are your thoughts on the research?

The post Research: AI mimics the human brain’s subconscious appeared first on AI News.

Feng Liu, Yong Shi, and Ying Liu put their test to AIs from Google, Apple, and Microsoft. The AIs scored as follows:

• Google: 47.28
• Baidu: 32.92
• Bing: 31.92
• Siri: 23.9

Google leads the pack comfortably according to the results. Baidu and Bing are a comfortable distance behind. Siri’s result is the most notable as it trails its competitors significantly with almost half the score of Google.

To calm those concerned about the potential AI apocalypse forewarned by the likes of Elon Musk — even Google’s score was relatively low compared to humans.

For comparison, the average score for a six-year-old in the researchers’ test is 55.5. For an 18-year-old, the average result is 96. This shows, at least for the time being, humans still have the advantage when it comes to general knowledge (…unless you’re under six, maybe.)

The gap is closing quickly, however. Earlier studies by the group show a rapid improvement over a relatively short period of time. Back in 2014, Google and Baidu put up scores of 26.5 and 23.5 respectively.

Once again, this posts Siri in a rather bad light. Google’s (and almost Baidu’s…) score back in 2014 was higher than Siri’s in the latest study.

The results are in-line with those found in another study posted by our sister publication IoT News. In that study, Siri also fell significantly behind Google and Microsoft in the ability to answer general knowledge questions.

The full research can be found here for your perusal.

What are your thoughts on the results of the study?

The post Google’s AI has a higher IQ than Siri, Bing, and Baidu — but all were beaten by a six-year-old appeared first on AI News.

Capgemini announced the findings into nearly 1,000 organisations today in its “Turning AI into concrete value: the successful implementer’s toolkit” study. 83 percent of the firms say the implementation of AI has generated new roles in their organisations while three-quarters have also seen an uplift in sales.

Further countering the idea that AI is destroying jobs; more than three in five (63%) of the organisations claim it has not resulted in any losses within their organisations.

Many organisations see artificial intelligence as a means to speed up tasks or automate mundane work for employees to spend time doing less routine or administrative tasks.

89 percent believe AI will make complex jobs easier and 88 percent think intelligent technologies will coexist with humans in their businesses. 71 percent have begun up-skilling or re-skilling employees in preparation.

“I think for every job that is lost, there will be many more jobs that are gained,” says the CTO of an unnamed large multinational corporation in the report. “The role of AI is not to replace humans, it is to augment humans. It is about helping us do what we do better.”

The research indicates that many organisations are yet to understand the full value potential offered by AI investments. 58 percent are focusing on ‘need to do’ implementations such as solving customer service issues, while just 46 percent are deploying ‘must do’ AI implementations with low complexity/high benefit. Firms tackling both simultaneously are set to gain the most.

In terms of regional adoption, Europe is falling behind. Countries such as Spain (31%), the Netherlands (24%) and France (21%) are further down the list in companies which have implemented AI. This is a stark comparison with countries such as India (58%) and Australia (49%).

You can find the full report here.

Are you concerned by the impact of AI on jobs? Share your thoughts in the comments.

The post Research: Despite popular opinion, AI is creating jobs appeared first on AI News.

The AI faced three of the best Dota 2 players in the world at the championships in the US and beat them all. It even learned the game from scratch using the old adage of “try, try, and try again.”

Some players of the eSports hit put in grueling hours of training for a chance to receive the increasingly large winnings on offer at some events. That means little to an AI, however, and it only cares about the glory.

The makers claim the goal ultimately is not just to win games and make human players look foolish, but to improve decision making.

“This is a step towards building AI systems which accomplish well-defined goals in messy, complicated situations involving real humans,” say the OpenAI engineers.

Should we bow to our AI overlords?

While the current crop of Dota 2 players appear unable to beat the AI, with enough commitment the creators believe it’s possible. The AI bots are not better in terms of actions-per-minute, but rather in their ability to make smarter decisions.

OpenAI can beat regular Dota 2 players in the space of an hour (which means it could learn to beat me in 30 minutes.) It takes approximately two weeks for the AI to beat the best Dota 2 players in the scene.

The AI wasn’t flawless, however. During subsequent tests against other attendees at the championship, some players were able to confuse the bot and distract it from the main objectives. Keep that in mind for the prophesied AIpocalypse.

Are you impressed with the decision-making of AIs such as OpenAI? Share your thoughts in the comments.

To learn more about Bot & Virtual Assistant Development attend the AI Conference and AI Exhibition in Silicon Valley this fall (29-30th November 2017).

The post OpenAI beats puny human players on Dota 2 appeared first on AI News.