If you’re handed a note that asks you to draw a picture of a bird with a yellow body, black wings and a short beak, chances are you’ll start with a rough outline of a bird, then glance back at the note, see the yellow part and reach for a yellow pen to fill in the body, read the note again and reach for a black pen to draw the wings and, after a final check, shorten the beak and define it with a reflective glint. Then, for good measure, you might sketch a tree branch where the bird rests.
Now, there’s a bot that can do that, too.
The new artificial intelligence technology under development in Microsoft’s research labs is programmed to pay close attention to individual words when generating images from caption-like text descriptions. This deliberate focus produced a nearly three-fold boost in image quality compared to the previous state-of-the-art technique for text-to-image generation, according to results on an industry standard test reported in a research paper posted on arXiv.org.
The technology, which the researchers simply call the drawing bot, can generate images of everything from ordinary pastoral scenes, such as grazing livestock, to the absurd, such as a floating double-decker bus. Each image contains details that are absent from the text descriptions, indicating that this artificial intelligence contains an artificial imagination.
“If you go to Bing and you search for a bird, you get a bird picture. But here, the pictures are created by the computer, pixel by pixel, from scratch,” said Xiaodong He, a principal researcher and research manager in the Deep Learning Technology Center at Microsoft’s research lab in Redmond, Washington. “These birds may not exist in the real world — they are just an aspect of our computer’s imagination of birds.”
The drawing bot closes a research circle around the intersection of computer vision and natural language processing that He and colleagues have explored for the past half-decade. They started with technology that automatically writes photo captions – the CaptionBot – and then moved to a technology that answers questions humans ask about images, such as the location or attributes of objects, which can be especially helpful for blind people.
These research efforts require training machine learning models to identify objects, interpret actions and converse in natural language.
“Now we want to use the text to generate the image,” said Qiuyuan Huang, a postdoctoral researcher in He’s group and a paper co-author. “So, it is a cycle.”
Image generation is a more challenging task than image captioning, added Pengchuan Zhang, an associate researcher on the team, because the process requires the drawing bot to imagine details that are not contained in the caption. “That means you need your machine learning algorithms running your artificial intelligence to imagine some missing parts of the images,” he said.
Attentive image generation
At the core of Microsoft’s drawing bot is a technology known as a Generative Adversarial Network, or GAN. The network consists of two machine learning models, one that generates images from text descriptions and another, known as a discriminator, that uses text descriptions to judge the authenticity of generated images. The generator attempts to get fake pictures past the discriminator; the discriminator never wants to be fooled. Working together, the discriminator pushes the generator toward perfection.
Microsoft’s drawing bot was trained on datasets that contain paired images and captions, which allow the models to learn how to match words to the visual representation of those words. The GAN, for example, learns to generate an image of a bird when a caption says bird and, likewise, learns what a picture of a bird should look like. “That is a fundamental reason why we believe a machine can learn,” said He.
GANs work well when generating images from simple text descriptions such as a blue bird or an evergreen tree, but the quality stagnates with more complex text descriptions such as a bird with a green crown, yellow wings and a red belly. That’s because the entire sentence serves as a single input to the generator. The detailed information of the description is lost. As a result, the generated image is a blurry greenish-yellowish-reddish bird instead a close, sharp match with the description.
As humans draw, we repeatedly refer to the text and pay close attention to the words that describe the region of the image we are drawing. To capture this human trait, the researchers created what they call an attentional GAN, or AttnGAN, that mathematically represents the human concept of attention. It does this by breaking up the input text into individual words and matching those words to specific regions of the image.
“Attention is a human concept; we use math to make attention computational,” explained He.
The model also learns what humans call commonsense from the training data, and it pulls on this learned notion to fill in details of images that are left to the imagination. For example, since many images of birds in the training data show birds sitting on tree branches, the AttnGAN usually draws birds sitting on branches unless the text specifies otherwise.
“From the data, the machine learning algorithm learns this commonsense where the bird should belong,” said Zhang. As a test, the team fed the drawing bot captions for absurd images, such as “a red double-decker bus is floating on a lake.” It generated a blurry, drippy image that resembles both a boat with two decks and a double-decker bus on a lake surrounded by mountains. The image suggests the bot had an internal struggle between knowing that boats float on lakes and the text specification of bus.
“We can control what we describe and see how the machine reacts,” explained He. “We can poke and test what the machine learned. The machine has some background learned commonsense, but it can still follow what you ask and maybe, sometimes, it seems a bit ridiculous.”
Text-to-image generation technology could find practical applications acting as a sort of sketch assistant to painters and interior designers, or as a tool for voice-activated photo refinement. With more computing power, He imagines the technology could generate animated films based on screenplays, augmenting the work that animated filmmakers do by removing some of the manual labor involved.
For now, the technology is imperfect. Close examination of images almost always reveals flaws, such as birds with blue beaks instead of black and fruit stands with mutant bananas. These flaws are a clear indication that a computer, not a human, created the images. Nevertheless, the quality of the AttnGAN images are a nearly three-fold improvement over the previous best-in-class GAN and serve as a milestone on the road toward a generic, human-like intelligence that augments human capabilities, according to He.
“For AI and humans to live in the same world, they have to have a way to interact with each other,” explained He. “And language and vision are the two most important modalities for humans and machines to interact with each other.”
In addition to Xiaodong He, Pengchuan Zhang and Qiuyuan Huang at Microsoft, collaborators include former Microsoft interns Tao Xu from Lehigh University and Zhe Gan from Duke University; and Han Zhang from Rutgers University and Xiaolei Huang from Lehigh University.
John Roach writes about Microsoft research and innovation. Follow him on Twitter.
I have decided to sell my Intel NUC DC3217BY due to lack of use.
If you’re not familiar with these, this is a fully featured Windows micro PC in a form factor no bigger than a typical android TV box.
It’s in excellent condition, still has the protective film on the top casing. Complete with original power supply, but sadly I no longer have the box it came in.
Intel Core i3-3217u
Windows 10 Pro x64
Here is a link to a review for more information:
Intel Next Unit of Computing (NUC – DC3217BY) Review
Price and currency: £125
Delivery: Delivery cost is included within my country
Payment method: PPG or BT
Advertised elsewhere?: Not advertised elsewhere
Prefer goods collected?: I have no preference
This message is automatically inserted in all classifieds forum threads.
By replying to this thread you agree to abide by the trading rules detailed here.
Please be advised, all buyers and sellers should satisfy themselves that the other party is genuine by providing the following via private conversation to each other after negotiations are complete and prior to dispatching goods and making payment:
- Landline telephone number. Make a call to check out the area code and number are correct, too
- Name and address including postcode
- Valid e-mail address
DO NOT proceed with a deal until you are completely satisfied with all details being correct. It’s in your best interest to check out these details yourself.
Sometimes big changes sneak up on you, especially when you’re talking about the future of data storage technology. For example, when exactly did full-on cloud adoption become fully accepted by all those risk-averse organizations, understaffed IT shops and disbelieving business executives? I’m not complaining, but the needle of cloud acceptance tilted over sometime in the recent past without much ado. It seems everyone has let go of their fear of cloud and hybrid operations as risky propositions. Instead, we’ve all come to accept the cloud as something that’s just done.
Sure, cloud was inevitable, but I’d still like to know why it finally happened now. Maybe it’s because IT consumers expect information technology will provide whatever they want on demand. Or maybe it’s because everything IT implements on premises now comes labeled as private cloud. Influential companies, such as IBM, Microsoft and Oracle, are happy to help ease folks formerly committed to private infrastructure toward hybrid architectures that happen to use their respective cloud services.
In any case, I’m disappointed I didn’t get my invitation to the “cloud finally happened” party. But having missed cloud’s big moment, I’m not going to let other obvious yet possibly transformative trends sneak past as they go mainstream with enterprises in 2018. So when it comes to the future of data storage technology, I’ll be watching the following:
- Containers arose out of a long-standing desire to find a better way to package applications. This year we should see enterprise-class container management reach maturity parity with virtual machine management — while not holding back any advantages containers have over VMs. Expect modern software-defined resources, such as storage, to be delivered mostly in containerized form. When combined with dynamic operational APIs, these resources will deliver highly flexible programmable infrastructures. This approach should enable vendors to package applications and their required infrastructure as units that can be redeployed — that is, blueprinted or specified in editable and versionable manifest files — enabling full environment and even data center-level cloud provisioning. Being able to deploy a data center on demand could completely transform disaster recovery, to name one use case.
- Everyone is talking about AI, but it is machine learning that’s slowly permeating through just about every facet of IT management. Although there’s a lot of hype, it’s worth figuring out how and where carefully applied machine learning could add significant value. Most machine learning is conceptually made up of advanced forms of pattern recognition. So think about where using the technology to automatically identify complex patterns would reduce time and effort. We expect the increasing availability of machine learning algorithms to give rise to new storage management processes. These algorithms can produce storage management processes that can learn and adjust operations and settings to optimize workload services, quickly identify and fix the root causes of abnormalities, and broker storage infrastructure and manage large-scale data to minimize cost.
- Management as a service (MaaS) is gaining traction, when looking at the future of data storage technology. First, every storage array seemingly comes with built-in call home support replete with management analytics and performance optimization. I predict that the interval for most remote vendor management services to quickly drop from today’s daily batch to five-minute streaming. I also expect cloud-hosted MaaS offerings are the way most shops will manage their increasingly hybrid architectures, and many will start to shift away from the burdens of on-premises management software. It does seem that all the big and even small management vendors are quickly ramping up MaaS versions of their offerings. For example, this fall, VMware rolled out several cloud management services that are basically online versions of familiar on-premises capabilities.
- More storage arrays now have in-cloud equivalents that can be easily replicated and failed over to if needed. Hewlett Packard Enterprise Cloud Volumes (Nimble); IBM Spectrum Virtualize; and Oracle cloud storage, which uses Oracle ZFS Storage Appliance internally, are a few notable examples. It seems counterproductive to require in-cloud storage to run the same or a similar storage OS as on-premises storage to achieve reliable hybrid operations. After all, a main point of a public cloud is that the end user shouldn’t have to care, and in most cases can’t even know, if the underlying infrastructure service is a physical machine, virtual image, temporary container service or something else.
However, there can be a lot of proprietary technology involved in optimizing complex, distributed storage activities, such as remote replication, delta snapshot syncing, metadata management, global policy enforcement and metadata indexing. When it comes to hybrid storage operations, there simply are no standards. Even the widely supported Amazon Web Services Simple Storage Service API for object storage isn’t actually a standard. I predict cloud-side storage wars will heat up, and we’ll see storage cloud sticker shock when organizations realize they have to pay both the storage vendor for an in-cloud instance and the cloud service provider for the platform.
- Despite the hype, nonvolatile memory express (NVMe) isn’t going to rock the storage world, given what I heard at VMworld and other fall shows. Yes, it could provide an incremental performance boost for those critical workloads that can never get enough, but it’s not going to be anywhere near as disruptive to the future of data storage technology as what NAND flash did to HDDs. Meanwhile, NVMe support will likely show up in most array lineups in 2018, eliminating any particular storage vendor advantage.
On the other hand, a bit farther out than 2018, expect new computing architectures, purpose-built around storage-class memory (SCM). Intel’s initial releases of its “storage” type of SCM — 3D XPoint deployed on PCIe cards and accessed using NVMe — could deliver a big performance boost. But I expect an even faster “memory” type of SCM, deployed adjacent to dynamic RAM, would be far more disruptive.
How did last year go by so fast? I don’t really know, but I’ve got my seatbelt fastened for what looks to be an even faster year ahead, speeding into the future of data storage technology.
This holiday season is all about giving. Whether you’re giving time to loved ones, giving gifts to family and friends, or giving your stomach more food than it can handle, Xbox wants to make it easy to play together and give together. From December 21 to January 4, you can join Xbox Game Pass and Xbox Live Gold for $1 each! Also, you can share your love of gaming by gifting an Xbox Game Pass membership.
Not only are we giving our fans great deals, we have a great opportunity for the Xbox community to give back to kids across the world. Xbox is partnering with GameChanger to make a difference for children in hospitals. For each Xbox Game Pass membership purchased or gifted from December 21 to January 4, Xbox will donate $10 of Xbox Game Pass to hospitals around the world through GameChanger.
Here at Xbox we believe in accessibility, diversity, and inclusion. We also believe that our passionate and awesome community of Xbox fans love to make a difference in people’s lives. That is why we are incredibly excited and humbled to team up with GameChanger this holiday to positively impact the lives of children facing life-threatening illnesses. Over the past several years, Microsoft has partnered with GameChanger and other organizations to place hundreds of thousands of Xbox consoles in hospitals around the world with the goal of providing entertainment to hospitalized children. Now we want to give our great community a chance to get involved as well.
Play 100+ Games for $1
Get your first month of Xbox Game Pass for $1 and spend your holiday with unlimited access to over 100 Xbox One and Xbox 360 games on the Xbox One family of devices. With new games added every month, there is always a new adventure waiting for you. All purchases benefit Season of Giving with Xbox. You must be signed in to get this offer (not available for existing members).
Go Gold for $1
Get your first month of Xbox Live Gold for $1! Go Gold for exclusive savings and join the best community of gamers on the most advanced multiplayer network. All purchases benefit Season of Giving with Xbox. You must be signed in to get this offer (not available for existing members).
Already own Xbox Game Pass or Xbox Live Gold?
Here’s how you can still get involved with the cause! Give the gift of 100+ games this holiday season by gifting Xbox Game Pass at full price. Also, you can add additional time to your Xbox Game Pass or Xbox Live Gold membership at full price.
For every Xbox Game Pass or Xbox Live Gold membership you buy or gift, Xbox will donate $10 of Xbox Game Pass to hospitals around the world through Season of Giving with Xbox.
Let’s play together & give together to bring joy to hospitalized children around the world. We hope we can make it a bit easier to make a difference this holiday season. Thank you for reading and have a wonderful end to the year.
If you’re just getting started with Hyper-V and struggling with the networking configuration, you are not alone. I (and others) have written a great deal of introductory material on the subject, but sometimes, that’s just too much. I’m going to try a different approach. Rather than a thorough deep-dive on the topic that tries to cover all of the concepts and how-to, I’m just going to show you what you’re trying to accomplish. Then, I can just link you to the necessary supporting information so that you can make it into reality.
First things first. If you have a solid handle on layer 2 and layer 3 concepts, that’s helpful. If you have experience networking Windows machines, that’s also helpful. If you come to Hyper-V from a different hypervisor, then that knowledge won’t transfer well. If you apply ESXi networking design patterns to Hyper-V, then you will create a jumbled mess that will never function correctly or perform adequately.
Your Goals for Hyper-V Networking
You have two very basic goals:
- Ensure that the management operating system can communicate on the network
- Ensure that virtual machines can communicate on the network
Any other goals that you bring to this endeavor are secondary, at best. If you have never done this before, don’t try to jump ahead to routing or anything else until you achieve these two basic goals.
Hyper-V Networking Rules
Understand what you must, can, and cannot do with Hyper-V networking:
What the Final Product Looks Like
It might help to have visualizations of correctly-configured Hyper-V virtual switches. I will only show images with a single physical adapter. You can use a team instead.
Networking for a Single Hyper-V Host, the Old Way
An old technique has survived from the pre-Hyper-V 2012 days. It uses a pair of physical adapters. One belongs to the management operating system. The other hosts a virtual switch that the virtual machines use. I don’t like this solution for a two adapter host. It leaves both the host and the virtual machines with a single point of failure. However, it could be useful if you have more than two adapters and create a team for the virtual machines to use. Either way, this design is perfectly viable whether I like it or not.
Networking for a Single Hyper-V Host, the New Way
With teaming, you can just join all of the physical adapters together and let it host a single virtual switch. Let the management operating system and all of the guests connect through it.
Networking for a Clustered Hyper-V Host
For a stand-alone Hyper-V host, the management operating system only requires one connection to the network. Clustered hosts benefit from multiple connections. Before teaming was directly supported, we used a lot of physical adapters to make that happen. Now we can just use one big team to handle our host and our guest traffic. That looks like this:
VLANs seem to have some special power to trip people up. A few things:
- The only purpose of a VLAN is to separate layer 2 (Ethernet) traffic.
- VLANs are not necessary to separate layer 3 (IP) networks. Many network administrators use VLANs to create walls around specific layer 3 networks, though. If that describes your network, you will need to design your Hyper-V hosts to match. If your physical network doesn’t use VLANs, then don’t worry about them on your Hyper-V hosts.
- Do not create one Hyper-V virtual switch per VLAN the way that you configure ESXi. Every Hyper-V virtual switch automatically supports untagged frames and VLANs 1-4096.
- Hyper-V does not have a “default” VLAN designation.
- Configure VLANs directly on virtual adapters, not on the virtual switch.
Other Quick Pointers
I’m going to provide you with some links so you can do some more reading and get some assistance with configuration. However, some quick things to point out:
- The Hyper-V virtual switch does not have an IP address of its own.
- You do not manage the Hyper-V virtual switch via an IP or management VLAN. You manage the Hyper-V virtual switch using tools in the management or a remote operating system (Hyper-V Manager, PowerShell, and WMI/CIM).
- Network connections for storage (iSCSI/SMB): Preferably, network connections for storage will use dedicated, unteamed physical adapters. If you can’t do that, then you can create dedicated virtual NICs in the management operating system
- Multiple virtual switches: Almost no one will ever need more than one virtual switch on a Hyper-V host. If you have VMware experience, especially do not create virtual switches just for VLANs.
- The virtual machines’ virtual network adapters connect directly to the virtual switch. You do not need anything in the management operating system to assist them. You don’t need a virtual adapter for the management operating system that has anything to do with the virtual machines.
- Turn off VMQ for every gigabit physical adapter that will host a virtual switch. If you team them, the logical team NIC will also have a VMQ setting that you need to disable.
For More Information
I only intend for this article to be a quick introduction to show you what you’re trying to accomplish. We have several articles to help you dive into the concepts and the necessary steps for configuration.
If you’re willing to post first that would be most appreciated.
If the total weight is less than 2kg then postage via special delivery should be £11 with plenty of insurance, which of course I’ll cover.
I’ll send you a pm shortly with my details for posting.
If you’re willing to post first that would be most appreciated.
If the total weight is less than 2kg then postage via special delivery should be £11 with plenty of insurance, which of course I’ll cover.
I’ll send you a pm shortly with my details for posting.
I’ve got a mid 2014 Macbook Pro Retina if you’re interested. It’s very clean, no marks, and has a low battery count of 198.
Its an i5, 8GB RAM with 256GB SSD storage.
It comes complete with charger and is boxed.
I’m looking for £575 plus postage for it (think thats about £25 but will need to check given location).
Let me know if interested and I’ll post photos.
Numbers don’t lie, and they don’t judge. They don’t care if you’re a woman, or where you grew up, or if you speak with an accent.
These are some of the thoughts I’ve had after speaking at the Grace Hopper Celebration of Women in Computing Conference and the Women in Statistics and Data Science Conference. After my presentations, I met with diverse groups of young women who are entering fields rooted in math and science. They wanted to know how I got here, what kind of training I’ve had, how I got on the management track, and how I came to lead a data analytics team at Microsoft.
I was honored and humbled by the experience. I was also proud, not about where I’ve come, but that the questions they are asking at the beginning of their careers are different than the ones I asked at the beginning of mine. Instead of wondering if they could pursue a career in data analytics, they asked me what steps they should take to become a leader in the data analytics field.
And while the questions are different now, and women and people of diverse backgrounds have more opportunities than when I was coming up, we have much more work to do. The Science, Technology, Engineering, and Math (STEM) field is still dominated by men, and even more so the senior or technical roles. For a variety of reasons, girls are not choosing to study these subjects in school as often as boys do, and when they do, they are still fighting for equality in many ways.
Yet despite all of this, I firmly believe the tide is turning, and that, if we keep pushing hard for more opportunities, there is reason for optimism. All I must do is look back at how I persisted in my journey from Calcutta, India, to Redmond, Washington, and then think back to the young people I met at Grace Hopper and Women in Statistics and Data Science to know this is true.
Making the numbers add up
Growing up in India in the 1960s and 70s, it was typical for more emphasis to be placed on boys—they were encouraged to go into STEM fields, whereas girls were typically expected to study humanities (literature, history, etc.). If you were lucky like me, and you had progressive parents who could afford private school, you absolutely went to an all-girls school where you could mostly pursue your career of choice (contending with some raised eyebrows if you leaned toward math and science). I attended all-girls schools from childhood all the way through getting my undergraduate degree in economics at Lady Brabourne College in Calcutta. It was only when I got my master’s degree in econometrics at Calcutta University that the classes I took were integrated with men.
You would think that attending a private school meant I would have a leg up on getting into a STEM field, but to me it felt like something I had to fight for every day. In secondary school, I remember we had only one math class and one science class per day, compared to plenty of writing, literature, and humanities classes. In primary school, my father always used to complain that we didn’t have enough math homework. At one of my parent-teacher conferences, he complained to the teacher, and the next day she gave us a bunch of extra math homework. The other students were like, “What’s with all the math?” My teacher looked at me, and said “you can thank Sarmila’s dad for that.”
This kind of subtle, limiting thinking continued throughout. When I was ready to graduate with my undergraduate degree in economics, I went to get a letter of recommendation for graduate school, and one of my professors actually blurted out, “When are you going to start a family?” Eventually, it worked out. I persisted enough to get my master’s degree, moved to the United States to get my doctorate, and launch myself into a career that eventually landed me at Microsoft.
Fly like an eagle
All my life, until I finished my Ph.D., I had a poster of an eagle flying alone in the twilight sky. That eagle was my talisman, always reminding me that, “In Life, Your Attitude Determines Your Altitude.” Working in tech in my era, I learned to tough it out, to hold on to my dreams, and to pick my battles. When things get challenging, I always think back to that eagle.
When I was talking to those young women at those conferences, I told them, “at times, even today, you’ll face stereotypical attitudes—you need to keep pushing for change. We all have come a long way on our circuitous journey. The tides are changing, and women are now ahead of where they used to be. But that said, firm conviction in yourself is your best ally.”
Tags: data analytics, women in IT