AI news

AI (artificial intelligence) is the simulation of human intelligence processes by machines, especially computer systems.

World latest news

What I learned when trying to improve an AI agent in a game using deep learning

Late 2018 I participated in kaggle’s “Quick, Draw! Doodle Recognition Challenge”.For those of you who are unaware, below is a short description of this game:“Quick, Draw!” was released as an experimental game to educate the public in a playful way about how AI works. The game prompts users to draw an image depicting a certain category, such as ”banana,” “table,” etc.As part of this competition, a subset of more than 1B drawings was released which had 340 labels. The competitors needed to improve the existing AI algorithm which distinguishes whether a user has correctly been able to draw what was asked for. For each test image, the need was to predict the three most probable classes the doodle might belong to.key_id,word9000003627287624,The_Eiffel_Tower airplane donut9000010688666847,The_Eiffel_Tower airplane donutThe finest algorithm was chosen based on its Mean Average Precision @ 3 (MAP@3).where U is the number of scored drawings in the test data, P(k) is the precision at cutoff kand n is the number of predictions per drawing.Initial deep dive into the dataThe drawings were captured as timestamped vectors, tagged with metadata including what the player was asked to draw and in which country the player was located. Each of the 340 classes had CSV files in the below format defining how each of the doodles was drawn by the corresponding player.A sample for doodles with apple as classUsing the below code we can convert each of the strokes in the drawing column of the above file into a corresponding image.BASE_SIZE = 256def draw_cv2(raw_strokes, size=299, lw=4, time_color=False): img = np.zeros((BASE_SIZE, BASE_SIZE), np.uint8) for t, stroke in enumerate(raw_strokes): for i in range(len(stroke[0]) - 1): color = 255 - min(t, 10) * 13 if time_color else 255 _ = cv2.line(img, (stroke[0][i], stroke[1][i]), (stroke[0][i + 1], stroke[1][i + 1]), color, lw) img = cv2.copyMakeBorder(img,4,4,4,4,cv2.BORDER_CONSTANT) if size != BASE_SIZE: return cv2.resize(img, (size, size), interpolation=cv2.INTER_LINEAR) else: return imgFor example here is one from the snowman file:snowmanWe converted all of the strokes into corresponding images and stored them in corresponding folders(train & test).A batch of images in training dataUsing a convolutional neural network to identify the doodleIdeally, there are multiple ways this problem could have been tackled, for example as there is a sequential component to it with strokes being a sequence of coordinates a recurrent neural network could also be used. I rather preferred to tackle this as a computer vision problem as it is more easier to test and learn by visualizing the results in an image problem than a sequential one like the one we are working on.The architecture chosen by us was Resnets and its variants(https://arxiv.org/abs/1512.03385). We started off with Resnet18 and gradually tested the problem for performance even across bigger architectures. Empirically I observed Resnet34 gave us more bang for each buck than any of other networks.What did I learn from initial experimentation from the dataA look at the data and subsequent runs suggested the need for this problem was a simplified network which could zoom through these large number of doodles. The need of the hour was a simplified network with the ability to run multiple epochs within a limited time frame. Hence, I did not even try any complicated architectures which in the end was a great decision.Using Resnet34 the highest volume of data that I could run my experimentation on was 30% and it did show that more data does help with the generalization ability when you have a simple but quite diverse(more number of labels) dataset.Noise in the dataOn further observation, it was observed that there was lots of noise in the training data, that is there were lots of doodles which were wrongly labeled. This was actually impacting the learning capability of the model as you are inherently giving wrong instructions to it. Possible solutions for this, which I could not try are the development of another network to identify wrongly labeled images or hand labeling high loss images(those where there is highest difference between actual and predicted).Where did I landThe highest MAP@3 I got was 0.91444 on the public leaderboard which generalized quite well with a score of 0.91318 on the private leaderboard. Considering the winner of the competition was on 0.95480 I was on the correct path. The one strategy that could have made a difference was if I had spent more time improving the noise in the data but this is a learning for next time.Thanks, everyone for reading my experience of tackling this extremely interesting problem! For anyone looking to try there hand at this below is the link to the competition.Quick, Draw! Doodle Recognition ChallengeWhat I learned when trying to improve an AI agent in a game using deep learning was originally published in Hacker Noon on Medium, where people are continuing the conversation by highlighting and responding to this story.
Hackernoon

China Telecom BestPay Raise 945mln to Focus on Blockchain and AI

China Telecom BestPay, the third largest payment platform operator in China, has raised 945 million yuan ($140 million) purposefully for bringing in blockchain and artificial intelligence advances. The company, founded in 2011, is a wholly owned subsidiary of state-owned China Telecom Corp (China’s third-largest wireless carrier) committed to mobile payment. Up to January 2016, its app called BestPay has attracted over 290 million users and become the third largest payment platform operator in China, closely following Alipay and WeChat Payment. Bestpay offers payment service in major countries including U.S., Australia, U.K., Japan, Korea, Canada and in Europe region, according to its website. According to local financial report, BestPay has sold 49 percent of its stakes to four renowned investors in exchange for 945 million yuan fund, which has been approved by the central bank. It is noteworthy that the company will use the raised money to seek improvement in big data and bring in two trending technologies – blockchain and artificial intelligence. Earlier last July, the country’s big three wireless carriers – China Mobile, China Unicom and China Telecom, have launched a blockchain research group for blockchain exploration. The company’s mixed ownership reform is a step to spin off the state-owned brand, which is also a boost to its vitality and further development, together with its strategy of adopting blockchain technology. It is evident that companies from varied fields are aggressively adopting blockchain around the whole world. Of late, international banking giant HSBC revealed that it has processed $250 billion of forex trades leveraging blockchain technology in 2018, though it may represent a small portion of the overall payment volumes for the bank. The figure is still impressive, showing the practice of blockchain uptake in the traditional banking sector. In China, apart from enterprises’ zeal for blockchain, the local government also attaches great importance to the nascent technology. As previously reported by 8btc, investment guided by the Chinese government in blockchain technology has stood at over 40 billion yuan (roughly $5.82 billion) this year alone. The number of blockchain patent applications from China shed more light on the trend. A total of 2,482 blockchain related patents are from China, and Chinese firms took 57 out of the top 100 global blockchain patent holders in 2018, with Alibaba and People’s Bank of China ranking among the top 5.
8BTC

Angelology and Artificial Intelligence

Human Minds in Relation to Non-Human MindsThe Assumption of the Virgin by Francesco Botticini.The contemporary discussion of artificial intelligence is not the first time human beings have wrestled with the possibility of other minds that are not human minds. Scholastic philosophers invested a considerable effort into understanding angels, with angels understood to be higher than human beings on the great chain of being. And the links of the great chain of being represent fundamentally different kinds of things that jointly constitute the world, and not a gradual ladder of progress, which is how it is often interpreted today.A significant chunk of Saint Thomas’ Summa Theologiae is devoted to angelology: Questions 50 through 64 is the “Treatise on Angels,” which goes into a luxury of detail in relation to the nature of angels. In the Bible, angels interact with human beings, but human beings are not the same kind of beings as angels, so this poses certain problems. Aquinas (and other Scholastics) takes the bull by the horns and wrestles directly with these problems posed by angels. Since angels do not need to breathe or drink or eat, which human beings must, there are questions that must be answered about, for example, what was going on when Lot invited two angels to his house in Sodom and he prepared them a meal.While these questions are interesting, much more interesting are the questions about the minds of angels — how they think, what they know, and how they know. Aquinas argues (First Part, Question: 54, Article: 5), citing Averroes to underline his argument, that angels are only intellect and will (and, by implication, that the angelic mind has no component of sensation) because they have no bodies naturally joined to them:“…In our soul there are certain powers whose operations are exercised by corporeal organs; such powers are acts of sundry parts of the body, as sight of the eye, and hearing of the ear. There are some other powers of the soul whose operations are not performed through bodily organs, as intellect and will: these are not acts of any parts of the body. Now the angels have no bodies naturally joined to them, as is manifest from what has been said already (Question [51], Article [1]). Hence of the soul’s powers only intellect and will can belong to them.” “The Commentator (Metaph. xii) says the same thing, namely, that the separated substances are divided into intellect and will. And it is in keeping with the order of the universe for the highest intellectual creature to be entirely intelligent; and not in part, as is our soul. For this reason the angels are called ‘intellects’ and ‘minds,’ as was said above (Article [3], ad 1).”For the contemporary reader coming from the mainstream of naturalistic Anglo-American analytical philosophy (meaning more-or-less my own philosophical perspective), the Thomist Treatise on Angels is a thought experiment that takes as its premise, “If there are beings such as this, what will the properties of these beings be?” When we today contemplate the possibility of strong AI (or, what I consider more interesting, machine consciousness), we are similarly asking, “If there are beings such as this, i.e., minds attributable to machines, what will these minds be like? How will they think? What will be their motives?”Would we say that machine minds have no corporeal bodies naturally joined to them, or would we say that the machine itself would be the corporeal body of some artificial intelligence? One of the persistent ideas about artificial intelligence is that this would be very different from human minds because of the lack of embodiment, and Aquinas seems to point to angels being very different from human beings not merely because they lack corporeal bodies, but because their lack of corporeal bodies means that the angelic mind is distinct from the human mind.One question that vexes artificial intelligence researchers, and especially those who speculate on superintelligence, is the orthogonality problem (the problems that arise from the orthogonality thesis, viz. that final goals and intelligence levels are independent). This problem does not exist for angelology, as we can be assured, ab initio, that the motivations of human beings and angels, along with all the rest of creation, are aligned with divine purpose, so that there cannot be a radical departure from the directionality to history imposed by an omniscient, omnipresent, and omnipotent deity.Now, it is true that rebellious human wills defied divine purpose, and indeed Satan defied divine purpose, but the universe is still unfolding according to the divine plan, despite these misbehaviors. To pursue this would take us deep into the theological permutation of the freewill/determinism debate (which, significantly, is not how the problem is posed or debated today), so we must be content with the idea that the Lord works in mysterious ways and leave it at that.In place of the theological framework within which our ancestors grappled with the problems of non-human minds, we have a naturalistic and scientific framework within which we think about the minds of machines, the minds of ETI, and the minds of our own species in the distant past and in the far future. In this naturalistic framework there are forms of directionality (though not that of a divine will) with which all existents must align, and by this I mean that arrows of time as they have been variously defined and distinguished.Different lists of arrows of time have been produced, all of which I have seen include the thermodynamic arrow of time, which thus points to some consensus. I consider the increasing metallicity of the visible universe to be as robust as the thermodynamic arrow of time, though I have not seen it any of these lists of arrows of time. However you choose to parse the irreversibility of the cosmos, if you recognize arrows of time that introduce irreversibility into the world, then this is a minimal naturalistic alignment that will hold for all minds, whether human, animal, ETI, machine, or otherwise. In other words, any or all of these minds would be minimally subject to the irreversibility of the arrow(s) of time.As I said, this is a minimal naturalistic alignment. I strongly suspect that an analysis of consciousness would reveal structures of consciousness that any consciousness would have in common with any other consciousness, and that would mean that there would be some structures of consciousness in common, as well as some structures only shared by a smaller subset of conscious agents. However, as I noted in Must a Philosophy of Mind be a Philosophy of Consciousness? and Extended Cognition and Naturalism, if a philosophy of mind is not also a philosophy of consciousness, we may have to treat the problems of mind and consciousness separately.If we do so, identifying a mind-body problem distinct from a consciousness-body problem, this distinction gives us a minimal framework for understanding machine intelligence that is not also machine consciousness (and, I suppose, vice versa). And there are some potential advantages to making this distinction. While the idea of extended cognition as applied to human beings or other biologically implemented minds seems, at times, a bit of a stretch, I would not hesitate nearly so much in attributing extended cognition to a machine mind. Every component that we plugged into a machine mind, or which we networked with a machine mind, would constitute a kind of extended cognition.Here we might get into philosophical problems unique to a machine mind that do no (at least, do not yet) apply to biologically implemented minds. For example, where would we draw the distinction between a machine mind proper and its extended cognition? If we plug a new module into a machine mind, is this simply a larger machine mind, or is it a machine mind with extended cognition? And there would be problems of a sorites paradox involved in making a machine mind larger or smaller: how many components can we take away from a machine mind and have it still be a mind?Angelology and Artificial Intelligence was originally published in Hacker Noon on Medium, where people are continuing the conversation by highlighting and responding to this story.
Hackernoon

CC Forum Malta: Blockchain, AI and Digital Innovation

CoinSpeaker CC Forum Malta: Blockchain, AI and Digital Innovation It will take place on 17-19 March 2019 in Malta connecting global thought leaders, policy makers, investors and startups from across the world for a 3 day top content event. It will be attended by the industry leaders, think tanks, institutional and private investors, family offices and VC firms. The Forum’s highlights include: 2500+ attendees 100+ influential speakers 20+ participants of the Investors’ Hub 50+ exhibititors To be inaugurated by Hon. Prime Minister, the forum is privileged to have some of the world’s most authoritative speakers, some of whom are global transformers: http://www.cc-forum.com/speakers/ Split across three tracks, the Forum’s agenda will address a wide range of issues including Blockchain and Foreign Direct Investment, the Future of Digital Investment and the Regulatory Framework of the Blockchain & Crypto Space. Part of the Forum’s programme are one-to-one duels where heavyweights will engage in heated public debates on the big issues of the space with the conference audience being interactively involved. The Forum will see an unprecedented agenda «The World’s Ecosystems and Crypto Investment» where a whole track will be given to crypto friendly governments who will be showcasing their ecosystems and highlighting their blockchain initiatives. Global announcements are expected to be made. A distinctive feature of CC Forum is the Investors’ Hub – an exclusive networking area where the brightest startups will have access to decision makers representing participating investment funds, VC firms and family offices, with a total of 70B USD under management. A series of high profile round tables involving Malta’s senior governmental officials and global investors. An ICO/STO contest will be held alongside the two day exhibition in the Hall’s lobby in which startups  are welcome to participate. Last, but not least, the Forum abounds in a rich networking programme ranging from postconference receptions to private VIP retreats. It will culminate in the Gala Dinner & Awards Giving Ceremony. CC Forum Malta: Blockchain, AI and Digital Innovation
Coinspeaker

The 1ST AI DApp Found on the Cortex Blockchain

After Cortex launched their AI-on-Blockchain TestNet in September 2018, they discovered a DApp developed by their community using Cortex technology. The Cortex team found the AI smart contract […] The post The 1ST AI DApp Found on the Cortex Blockchain appeared first on UseTheBitcoin.
Use The Bitcoin
More news sources

AI news by Finrazor

FACT

Have you heard about pilotless vehicles on blockchain?

As a byproduct of this development, the partnership between Boeing and SparkCognition 'will also provide a standardized programming interface to support package delivery, industrial inspection and other commercial applications', according to Boeing press release

Trending

Hot news

Hot world news

Stellar Price Analysis: Grayscale Announces XLM Based Trust; XLM/USD Stuck Within Bearish Structure

Global digital asset management firm, Grayscale, has announced an investment vehicle based around XLM. XLM/USD is moving within the confinements of a bearish pattern structure, subject to a breakout south. XLM/USD has been subject to very narrow and choppy trading, which has been going on for the past eight sessions now. Price action is moving […] The post Stellar Price Analysis: Grayscale Announces XLM Based Trust; XLM/USD Stuck Within Bearish Structure appeared first on Hacked: Hacking Finance.
Hacked

Grayscale Adds Stellar as Latest Cryptocurrency Investment Trust

Grayscale Adds Stellar as Latest Cryptocurrency Investment Trust Digital currency investment group Grayscale confirmed it had successfully launched its latest fund, dedicated to Stellar’s Lumens (XLM) token, in a tweet Jan. 17. Grayscale, which now operates nine cryptocurrency funds, timed the move to coincide with a change of image for its products, renaming all its […] Cet article Grayscale Adds Stellar as Latest Cryptocurrency Investment Trust est apparu en premier sur Bitcoin Central.
Bitcoin Central

Researches from MIT, Stanford Set to Replace Bitcoin with Their Groundbreaking Crypto Project

CoinSpeaker Researches from MIT, Stanford Set to Replace Bitcoin with Their Groundbreaking Crypto Project Until now, everybody has been talking about Bitcoin, the most popular and widely used digital currency. However, Bitcoin is unable to process thousands of transactions a second. Researchers from the Massachusetts Institute of Technology (MIT), UC-Berkeley, Stanford University, Carnegie Mellon University, University of Southern California, and the University of Washington have decided to fix such a weakness and develop a crypto asset better than Bitcoin. The researchers are working together as Distributed Technology Research (DTR), a non-profit organization based in Switzerland and backed by hedge fund Pantera Capital. The first initiative of Distributed Technology Research is the Unit-e, a virtual coin that is expected to solve bitcoin’s scalability issues while holding true to a decentralized model and process transactions faster than even Visa or Mastercard. Babak Dastmaltschi, Chairman of the DTR Foundation Council, said: “The blockchain and digital currency markets are at an interesting crossroads, reminiscent of the inflection points reached when industries such as telecom and the internet were coming of age. These are transformative times. We are nearing the point where every person in the world is connected together. Advancements in distributed technologies will enable open networks, avoiding the need for centralized authorities. DTR was formed with the goal of enabling and supporting this revolution, and it is in this vein that we unveil Unit-e.” According to the press release, Unit-e will be able to process 10,000 transactions per second. That’s worlds away from the current average of between 3.3 and 7 transactions per second for Bitcoin and 10 to 30 transactions for Ethereum. Joey Krug, a member of the DTR Foundation Council and Co-Chief Investment Officer at Pantera Capital, believes that a lack of scalability is holding back cryptocurrency mass adoption. He said: “We are on the cusp of something where if this doesn’t scale relatively soon, it may be relegated to ideas that were nice but didn’t work in practice: more like 3D printing than the internet.” The project’s ideology is firmly rooted in transparency, with a belief in open-source, decentralized software developed in the public interest with inclusive decision-making. The core team of the project is based in Berlin. To solve the scalability problem, DTR has decided to develop the Unit-e with parameters very close to Bitcoin’s design, but many things will be improved. Gulia Fanti, DTR lead researcher and Assistant Professor of Electrical and Computer Engineering at Carnegie Mellon University, commented: “In the 10 years since Bitcoin first emerged, blockchains have developed from a novel idea to a field of academic research. Our approach is to first understand fundamental limits on blockchain performance, then to develop solutions that operate as close to these limits as possible, with results that are provable within a rigorous theoretical framework.” The launch of the Unit-e is planned for the second half of 2019. Researches from MIT, Stanford Set to Replace Bitcoin with Their Groundbreaking Crypto Project
Coinspeaker

BitPay CEO Says Bitcoin Is Solving Real Problems Around the World

BitPay co-founder and CEO, Stephen Pair, has recently commented that Bitcoin (BTC) is solving several issues around the world. He said that in a press release uploaded a […] The post BitPay CEO Says Bitcoin Is Solving Real Problems Around the World appeared first on UseTheBitcoin.
Use The Bitcoin
By continuing to browse, you agree to the use of cookies. Read Privacy Policy to know more or withdraw your consent.