SB Infowaves- the leader of future AI software development
New Delhi [India], May 4: In a significant development in India’s fast-evolving technology services sector, SB Infowaves, under the leadersh
In a significant development in India’s fast-evolving technology services sector, SB Infowaves, under the leadership of founder and Managing Director Shreya Parasrampuria, is strengthening its global footprint through strategic partnerships, advanced AI-driven offerings, and active participation in major industry platforms.
Founded with a vision to deliver practical, scalable technology solutions, SB Infowaves has grown into a multi-domain digital transformation company offering services across artificial intelligence, machine learning, custom software development, cloud infrastructure, web and mobile applications. The company has positioned itself strongly in emerging areas such as AIoT (Artificial Intelligence of Things), integrating connected devices with intelligent data systems to enable real-time decision-making and automation for businesses across industries.
As the MD, Parasrampuria brings a great combination of finance and strategy, having gained much experience working in PwC and ICICI Bank. She places an emphasis on solving problems and implementing them by creating unique technological platforms. Thanks to her guidance, SB Infowaves managed to increase its delivery across the globe, focusing on the client-centeredness and tangible results.
SB Infowaves is planning to take part in CMPL Expo 2026 that is going to be held between May 4 and June 6, 2026, at Jio World Convention Centre in Mumbai. Being considered as one of the most prominent events for the contract manufacturing and private label sector in Asia, the event is going to be attended by such companies as those involved in the FMCG industry.
The involvement of the firm in the CMPL Expo showcases its attempts at expansion into various sectors such as retail, manufacturing, and e-commerce, where there is an increasing demand for AI-based automation, data analytics, and digitization.
It should be highlighted that SB Infowaves has formed an international alliance with X-Venture,US with the intention of delivering innovative Agentic AI services to small firms in the US market. Through this international collaboration, companies would be able to deploy autonomous AI systems that not only perform automation but also take decisions.
By becoming ISO 27001:2022 compliant, along with its wide range of offerings, such as Artificial Intelligence (AI), Blockchain, Cloud Computing, DevOps, and Data Engineering, SB Infowaves will continue with its position as a reliable technology partner for companies that seek futuristic digital environments.
In the ten years he had been a vigilante, Peter Parker had become very good at sneaking into places he shouldn't have been.
Air vents were useless. The vast majority of them were far too narrow for anybody to slip through without becoming amorphous, and even when the ducts were large enough it was impossible to move inside one without making enough noise to alert the entire block. The subceiling--the space above the ceiling tiles, but below the actual architectural structure--was a far better bet, but that was similarly cramped--and besides, only some buildings had gaps in the walls to allow for movement like that.
Using a disguise to sneak in was better all around, but it required a lot of skill and care. You couldn't disguise yourself as a scientist unless you were genuinely an expert in the field you were pretending to study. Nor, in this particular case, could you just dress up as any old agent--they had security levels. Executives were out, reporters were only viable if the people you were trying to fool had reason to believe a reporter was going to be there, and the less said about solicitors the better. The key was to attract as little scrutiny as possible, to not raise any questions you'd have trouble answering; because the second someone grew suspicious of you, your cover was all but blown.
Janitors, then, were perfect.
Nobody pays attention to a janitor. It's practically one of the perks of the job. Beneath notice means beneath scrutiny, and people only give custodians the slightest thought when a place needs cleaned. Even then it's just an assertion that a custodian needs to be there. Nobody questions what a janitor is doing in a room, even in the dead of night. Nobody questions why a janitor is wearing gloves, or where they got their ring of keys. There's no better disguise for going somewhere that people generally can't go.
Peter had been pretending to be a janitor in the main headquarters of the Cape Code Authority for several days now. He had listened intently as he'd mopped the floors, mapped out the layout in his head, figured out where the labs were and who had access to what while keeping his head down. He'd owned this coverall for years now, for infiltrations exactly like this, and now with the security cameras disabled he hooked his cart on the handle of his mop and dragged it towards the door the three agents had just left.
The door had locked automatically. Of course it had, all laboratory doors locked automatically around here, and even the custodians needed special permissions to get them unlocked. But as the door had swung closed, Peter had pressed the trigger in his palm under the guise of adjusting his grip on his mop, and now the door's latch was glued down beneath a small splatter of webbing. Pulling on a latex glove, Peter tugged the door open a crack and slipped into the lab.
He adjusted his hat as he glanced around the lab, the hat that had blond curls sewn to the inside to disguise his brown hair, and scratched at his false nose. The hologram table sat in the center of the room, still softly glowing even after its deactivation--an enormous waste of energy, but apparently nobody cared. Ignoring it entirely, Peter headed straight for the computer monitors against the far wall, grabbing a chair without breaking stride and only stopping to climb on top of it and crouch on the seat like a gargoyle on a rooftop.
Like everything Reed Richards ever touched, the computers were encrypted. But Peter had dated Johnny Storm for five months once, and he didn't spend so much time nearby his fellow supergenius without taking some time to figure out how to bypass their usual security. It took him just over five minutes to get through the firewalls, and then he stuck a translucent plastic sticky note to the screen and began to browse.
The sticky note was, of course, a data drive. Peter had learned about these only recently, but he was fast growing to like them; they were easy to conceal on his person and, unlike a USB stick, didn't require a specific size of port. As he opened up the computer's files, the drive pinged off of the computer's software and integrated itself into the system without leaving a trace. Cracking his knuckles, Peter typed a few cursory searches into the file browser and tapped Enter.
Perpetual Holographic Avatar/Nano-Tech Offensive Monsters had been a thorn in his side for over two years now. They didn't move like humans; their range of motion didn't have the limits that their skeletal shape implied; their systems adapted and learned and coordinated in ways that he'd never seen before in artificial intelligence. Even Octavius, permanently on the cutting edge of AI and biorobotics development, wasn't sure what the hell was going on with them. Last year, in the middle of beating the multi-armed megalomaniac's face in, Spider-Man had asked for Otto's thoughts on the Phantoms; the technology, both of them suspected, wasn't exactly beyond Otto's work so much as to the side of it. The systems were hyperspecialized: they had no connection to neural networks of old, and were practically useless for advancing them in the future. They were, in a word, alien.
Peter suspected Chitauri tech. The War of the Worlds had left countless remnants of the Chitauri on Earth; some of them still remained, like the Leviathan rotting in Maine, but far too many of them had seemed to simply vanish. Anyone who gave it more than ten seconds of thought could realize that governments of the world had squirreled the stuff away to study and reverse-engineer. Now, as Peter's eyes darted back and forth across the screen, he skimmed through the blueprints and models that he found in the folder and tried to see if any of it matched the distinctive look of the Chitauri.
Some of it did, he found as he kept searching, but not a huge amount. Reed had done some work with Chitauri tech in the past; traces of its influence were obvious in the composition of the Phantoms' gun barrels, and in the way their hard-light armor projected itself over the skeleton. Kid stuff, nothing that explained the problems he'd had with them. Peter's brow furrowed as he copied the files he found to his data drive and peered over his shoulder at the hologram table behind him.
What had Reed been saying to Flint in here only a few minutes ago? Peter had a spiderlike hypersensitivity to vibration; he could feel footsteps on the other side of the building rumbling through the floor, and the variations in air pressure caused by the fly drifting around the ceiling. But it didn't work like hearing did, nor was it interpreted by the same part of the brain. Though he had felt Reed talking in here, it just felt like a continuous drone of vibration against his skin--he hadn't heard him, and so couldn't interpret the words. And, like an idiot, he hadn't thought to bug the room beforehand.
He pushed his tongue against his upper lip in thought. Had it had something to do with why Flint had registered with the CCA in the first place?
Kicking a foot against the bottom of the desk, Peter rolled his chair over to the hologram table and set to work getting past the security there too. This took even less time than it had with the computer, now that Peter knew how Reed had updated his security measures over the last few years. Within three minutes of typing so fast an observer would have seen his fingers as blurs he was browsing through the most recently accessed files.
The image lifted out of the table and filled the room with its soft light, and Peter frowned at the image of the Phantom he saw. How on earth was this related to Flint's desire to Be A Real Boy? He typed a few commands into the table and watched the Phantom's white shell disappear to reveal the mechanical skeleton beneath. A few notes by Reed appeared to highlight key points, and Peter read through each with steadily rising concern.
Very little of the Sandman's mass was actually Flint Marko. When he had been disintegrated all those years ago, most of his body had become just plain old sand--only his nervous system had become anything different. Over the years, he had gained entire truckloads worth of sand and lost enough to fill beaches, but the gallon or so of milky white silica that had once been his brain and nerve cells had remained, scattered evenly through every shape and sculpture he made himself into. They assimilated granules of a similar composition through static cling, arranging them with an intricate electric charge that neither Flint nor Peter had ever fully understood, and now it looked like Reed wanted to apply that same static charge to the Phantom project.
Looking through the notes, Peter could already see that Reed wasn't putting much effort into following through on his promise. The conjectures and theories put forth in them were ludicrous--ideas that Peter had discarded years ago in his various scrambles to stop one of Marko's rampages. But he read between the lines, read ideas that Reed had intended for his own eyes only, and his blood grew steadily colder in his veins.
It wouldn't take much modification to turn a Phantom into a suitable chassis for Flint's nerve granules, so went Reed's idea. The skeleton already contained organic elements, and they already received commands from a biological source rather than a computer. This flew in the face of Peter's assumptions about the Phantoms.
They were only partially robots. They were like Octobots; their processing units were very much alive.
Peter waved a shaky hand over the table. The hologram deactivated, which wasn't his intent at all, but he was too taken aback to care.
Deep in the bowels of the building, ignored by Peter until now but always scratching at the back of his mind, the vibrations of mechanical footsteps rumbled through the walls and floor. The central hub of manufacturing and deploying Phantoms was located fifty feet under the foundation--a fact he'd known all along, but which he had to investigate now. Now, when he knew that within those robotic skeletons were living and thinking beings. Now, when he knew that the drills whirring and 3D printing that he felt even from here were working tirelessly to imprison and enslave something. Jumping off the chair, he retrieved his data drive from the computer and took barely a minute to wipe all evidence of his presence from the room. Then, readjusting his disguise and checking for the presence of witnesses, he slipped out of the room and finally allowed the door to lock.
The route to the underground hub was a circuitous one. As the operations were almost entirely automated, not even the janitors were given clearance to enter that level; maybe four people had access, and Peter wasn't one of them. No matter. There were more ways to sneak around than just throwing on a coverall and mopping a floor. If Peter's disguise only took him this far and no farther, it was time to drop it. Some places you could only reach as the wall-crawler.
Had the security cameras not mysteriously lost power earlier that afternoon, they would've seen a janitor shedding his hat, kicking off his shoes, and beginning to unbutton his coverall. Without breaking stride, he snatched a small bag from where he'd hidden it in his cart before and pulled on a mask; whatever features, real or fake, a witness might have noticed, they were now hidden by dark red fabric and two gleaming grey bug eyes. In short order the coverall and hat were gone--wrapped up into a web-knapsack that he slung onto his back even as he swapped his shoes out for red spandex boots. Pulling on his gloves right as he reached the elevator, Spider-Man stopped to politely tap the call button beside the sliding metal doors.
With a ding, the elevator doors slid open, and Spider-Man immediately smashed straight through the emergency hatch at the top of the lift.
Elevator shafts were always a bit more complicated than one expected. Even Peter, who could feel the constant motion of the metal boxes through the building and their cables sliding against pullies, always needed a moment to figure out how to squeeze through the systems that controlled its rise and fall. He paused as he examined the mechanism of this particular elevator before he sucked in his stomach and crawled around the box with a few inches to spare. Then, once he was beneath it, he released his grip on the elevator shaft and let himself fall.
He caught himself fifty feet later, his fingertips sticking instantly to the concrete as he touched it. Just across the shaft from him was a set of elevator doors, which he hopped onto and began to pry apart. It was slow going. Like everything in the CCA headquarters, these doors were made with superhumans in mind, and they had a magnetic lock that Spider-Man found himself straining to overpower. He pulled on them for a few seconds, changed his mind, and crawled two feet to the left to begin messing with the wiring that controlled the lock. There was a moment of silence, a low, hollow ding, and the doors slid open.
With one hand still stuck to the wall Spider-Man lowered himself into the unlit chamber, dropping to the floor and landing there in a crouch. What little light had made it down with him reflected off his mask's glaring eyes. For a moment he was still, one hand pressed to the metal beneath him and his attention fully on the vibrations of the environment. Then, mentally sorting through the sea of threats that his spider-sense whispered and squirmed at, he rose to his feet and nonchalantly slapped the lightswitch on the wall behind him. Sparse florescent lights flickered on above him, and he blinked and furrowed his brow as he adjusted.
Now that he was down here the vibrations were sharper, like a the world coming into focus as you come up from underwater. They travelled through the air, through the concrete, and through a metal catwalk that served as a floor, branching into pathways and situated above buzzing, whirring machinery. No wonder it had been so difficult to discern what was going on up above, Spider-Man reflected as he glanced over the guardrail and watched robotic limbs carry a Phantom chassis through a gap in the wall and to another room. He turned his attention ahead of him, where similar chasses were held in racks upon racks that spanned nearly wall to wall across the room, black robotic skeletons awaiting deployment.
But there was a difference between these Phantoms and the ones he so often encountered on the battlefield. Frowning under the mask, Spider-Man stepped forward, leaned over the catwalk's railing, and set a finger against the nearest collection of servos and solid-light projectors. Yes. There it was, the constant, ambient tremor of air in motion; the chasses were hollow like the frame of a bicycle. Whenever he'd fought them, they hadn't displayed any such emptiness.
Right. Mechanical systems supported by biological processing. He took his attention away from the chasses, looking instead at that hole in the wall that one of them had vanished into as he'd come in here. He could feel the Phantom in the next room over being hooked up to--to something, metal vibrating on contact with metal and stabilizing with a little pop. His eyes narrowed. His fingers twitching nervously, his breath held, he began to pace down the catwalk towards the door to that room.
A window on one side greeted him as he stepped through, displaying the Phantom under maintenance. Screens embedded into the window offered diagnostics and schematics, all of which Spider-Man ignored. He turned instead to the far wall, where what looked like a large cabinet was anchored in place and had a hundred or so pipes no wider than test tubes leading into and out of it. A quick ripping of metal, and he tossed a mangled padlock over his shoulder as he threw the cabinet doors open. The interior was poorly organized, and called to mind a prototype rather than anything intended for widespread implementation: a screen with a series of codes flashing across it, a mess of piping and tubing, and in carefully arranged racks hundreds upon hundreds of test tubes, most full of some amorphous fluid.
Spider-Man's brow furrowed as he selected a vial at random. Working carefully, he unscrewed the valve that connected it to the mess of piping and slid it out of the vial's stopper--without it, the test tube's lid sealed airtight again. He held it above eye level and turned to see the light filter through from overhead. The fluid inside surrounded what looked almost like a pipe cleaner, thousands of copper wires branching out from a central silicon rod. As he tilted it one way, an air bubble slid up the glass wall, and out of the corner of his eye he thought he saw--
--a tendril, as black as the rest of the liquid, squirming in that air pocket in a bid for freedom.
Spider-Man's eyes widened behind the mask. Oh my god.
Dead Leviathans and alien technology hadn't been the only things the Chitauri had brought to Earth. It had taken the terrestrial armies, and the remnants of SHIELD that Spider-Man had fought alongside, far too long to realize that the shape-shifting battlesuits that their enemies had used were themselves a separate species. Earth hadn't been the only planet to face invasion under the Chitauri; centuries ago, those invaders had conquered and enslaved a species called Klyntar. Amorphous, shapeshifting, symbiotic creatures, the Klyntar had the distinction of being able to use every single cell as musculature, digestive system, armoring, and neurons. Nobody was sure how long the Chitauri had been selectively breeding and brainwashing their symbiote slaves into battle armor, and until now Spider-Man had assumed that practice had stopped with the aliens' defeat.
The little vial of Klyntar sample in his hand was far from his first experience with the species. He had, for six months during and after the war, worn a stolen symbiote as a battlesuit of his own, and even after he and Vee had separated he'd been up close and personal with the species many, many times. But he had believed that Vee's defection from the Chitauri had been a fluke; that they had been the only Klyntar to be recovered from the Chitarui alive.
But now Spider-Man stood in the basement of the Cape Code Authority, holding a vial that contained another member of that species, and right next to him were over a hundred identical vials. All at once, the control systems of the Phantoms became obvious to him.
Without hesitation he turned back to the cabinet and began yanking the tubes out of their holders. The brush-like machinery in each vial, he figured as he worked, must have been some kind of brainwashing system; the copper wires made contact with as many of the Klyntar's neurons as possible, with controlled electric shocks frying out whatever thoughts the aliens tried to form and replacing them with--with whatever programming was necessary to get the Phantoms working. As he pulled each tube out, he killed the electrical charge, but for now he didn't release the Klyntar within from their cells. Where would they go down here? Did they even remember what they were? At best they'd die, at worst the CCA would collect them again and make it even harder to get to them again. No, for now he stuck the vials together with webbing, bundling them together in a padded sack of sorts--he could keep them safe until he knew what else to do, but for now--
--for now, he could feel footsteps vibrating through the concrete fifty feet above. Could feel the elevator starting to move, and the frantic tingling in his head suddenly concentrated all its alarm on the man upstairs. He paused, but only for the smallest fraction of a second; then he worked even faster, his hands becoming blurs again. Grab, break, thwip, grab, break, thwip. The bundle of vials and webbing in his arms was becoming untenably large. He kept at it anyway, always careful not to smash the vials, always careful to separate them from their neighbors with a carefully padded layer of webbing. Even as he webbed up the last one, he wove backpack straps onto the sack and pulled them onto his shoulders. Then he turned on his heel and darted out the door, ready to make an escape.
But as the elevator began its slow descent towards him, he paced around the room and realized that there was no escape to be found. No windows or doors, because he was in a basement, and the air ducts were of course far too small to crawl through. If he didn't have the Klyntar vials, he would've been able to crawl past the elevator, but with that bundle on his back there was no room. If he wanted to save these Klyntar, he was trapped down here with them.
Well, decided Spider-Man as his pacing came to a stop directly in front of the elevator. If he was about to be discovered down here, he certainly wasn't going to let whoever was about to discover him get a dramatic moment about it. There would be no voice booming out from behind him as he frantically looked for a hiding place, there would be no cat and mouse as the person looked for him in this increasingly exposed room. He folded his arms and leaned against the guardrail right in front of the elevator, glaring at the doors. Waiting.
When the doors dinged open, Scrier momentarily hesitated, not having expected to see Spider-Man so out in the open. He blinked behind those blank white eyes, far more awkward than a supervillain wanted to be, before he lamely managed, "I thought that was you, Spider-Man."
Five 2016 Trends We Expect to Come to Fruition in 2017
The start of a new year is an excellent occasion for audacious extrapolation. Based on 2016 developments, what do we expect for 2017?
This blog post covers five prominent trends: Deep Learning Beyond Cats, Chat Bots - Take Two, All the News In The World - Turning Text Into Action, The Proliferation of Data Roles, and What Are You Doing to My Data?
(1) Deep Learning Beyond Cats
In 2012, Google found cats on the internet using deep neural networks. With a strange sense of nostalgia, the post reminds us how far we have come in only four years, with more nuanced reporting as well as technical progress. The 2012 paper predicted the findings could be useful in the development of speech and image recognition software, including translation services. In 2016, Google’s WaveNet can generate human speech, General Adversarial Networks (GANs), Plug & Play Generative Networks, and PixelCNN can generate images of (almost) naturalistic scenes including animals and objects, and machine translation has improved significantly. Welcome to the future!
In 2016, we saw neural networks combined with reinforcement learning (i.e., deep reinforcement learning) beat the reigning champion Lee Sedol in Go (the battle continues online) and solve a real problem; deep reinforcement learning significantly reduces Google’s energy consumption. The combination of neural networks with probabilistic programming (i.e., Bayesian Deep Learning) and symbolic reasoning proved (almost) equally powerful. We saw significant advances in neural network architecture, for example, the addition of long-term memory (Neural Turing Machines) which adds a capacity resembling “common sense” to neural networks and may help us build (more) sophisticated dialogue agents.
In 2017, enabled by open-sourced software like Google’s TensorFlow released in late 2015, Theano, and Keras, neural networks will find (more) applications in industry (e.g., recommender systems), but widespread adoption won’t come easily. Algorithms are good at playing games like Go because games easily allow to generate the amount of data needed to train these advanced, data-hungry algorithms. The availability of data, or lack thereof, is a real bottleneck. Efforts to use pre-trained models for novel tasks using transfer learning (i.e., using what you have learned on one task to solve another, novel task) will mature and unlock a bigger class of use cases.
Parallel work on deep neural network architecture will enhance said architecture, deepen our understanding, and hopefully help us develop principled approaches for choosing the right architecture (there are many) for tasks beyond “CNNs are good for translation invariance and RNNs for sequences”.
In 2017, neural networks will go beyond game playing and deliver on their promise to industry.
(2) Chat Bots - Take Two
2016 had been declared by many the year of the bots, and it wasn’t. The narrative was loud but the results, more often than not, disappointing. Why?
Amongst the many reasons; lack of avenues for distribution, lack of enabling technologies, and the tendency to treat bots as a purely technical not product or design challenge. Through hard work and often failure, the best driver of future success, the bot community learned some valuable lessons in 2016. Bots can be brand ambassadors (e.g., Casper’s Insomnobot-3000) or marketing tools (e.g., Call of Duty’s Lt Reyes). Bots are good for tasks with clear objectives (e.g., scheduling a meeting) while exploration, especially if the content can be visualized, is better left to apps (you can, of course, squeeze it into a chatbot solution). Facebook’s messenger platform added an avenue for distribution; Google (Home, Allo) may follow while Apple (Siri) will probably stay closed. Facebook’s Wit.ai adds technology to enable developers to build bots, at re:Invent 2016, Amazon unveiled Lex.
After excitement and inflated expectations in 2016, we will see useful, goal-oriented, narrow-domain chatbots with use case appropriate personality supported by human agents when the bot’s intent recognition fails or when it wrangles a conversation. We will see more sophisticated intent recognition, graceful error handling, and more variety in the largely human-written template responses while ongoing research into end-to-end dialogue systems promises more sophisticated chatbots in the years to come. After the hype, a small, committed core remains and they will deliver useful chatbots in 2017.
Who wins our “The Weirdest Bot Of 2016” award? The Invisible Boyfriend.
(3) All the News In the World - Turning Text into Action
At the beginning there was the number; algorithms work on numerical data. Traditionally, natural language was difficult to turn into numbers that capture the meaning of words. Conventional bag-of-word approaches, useful in practice, fail to use syntactic information and fail to understand that “great” and “awesome” or Cassius Clay and Muhammad Ali are related concepts.
In 2013, Tomas Mikolov proposed a fast and efficient way to train word embeddings. A word embedding is a numerical representation of a word, say “great”, that is learned by studying the context in which the “great” tends to appear. The word embedding captures the meaning of “great” in the sense that “great” and “awesome” will be close to one another in the multi-dimensional word embedding space, the algorithm learned they are related. Alternatives like GloVe, word2vec for documents (i.e., doc2vec), and underlying methods like skip-gram and skip-thought further improved our ability to turn text into numbers and opened up natural language to machine learning and artificial intelligence.
In 2016, fastText allowed us to deal with out-of-vocab words (words the language model was not originally trained on) and SyntaxNet enhances our ability to not only encode the meaning of words but to parse the syntactic structure of sentences. Powerful, open-source natural language processing tool kits like spaCy allow data scientists and machine learning engineers without deep expertise in natural language processing to get started. FastText? Just pip install! Fuelled by this progress in the field, we saw a quiet but strong trend in industry towards utilizing these new powerful natural language processing tools to build large-scale industry applications that turn 6.6M news articles into a numerical indicator for banking distress or use 3M news articles to assess systemic risk in the European banking system. Algorithms will help us not only to make sense of the information in the world, they will help write content, too, and of course they will help bring our chit chatty chat bots to live.
In 2017, we will expect more data products built on top of vast amounts of news data especially data products that condense information into small, meaningful, actionable insights. Our world has become overwhelming, there is too much content. Algorithms can help! Somewhat ironically, we will also be using machines to create more content. A battle of machines.
In a world shaken by “fake news”, of course, one may regard these innovations with suspicion. As new technology enters the mainstream there is always hesitancy, but the critics are right. How do we know the compression is not biased? How do we train people to evaluate the trustworthiness of the information they consume especially when it has been condensed and computer generated? How do we fix the incentive problem of the news industry, distribution platforms like Facebook do not incentivise for deep, thoughtful writing; they monetize a few seconds of attention and are likely to feed existing biases, not all challenges are technical but should concern technically minded people.
The best AI writer of the year goes to? Benjamin, a recurrent neural network that wrote the movie script Sunspring.
(4) The Proliferation Of Data Roles
Remember when data scientist was branded the sexiest job of the 21st century? How about machine learning engineer? AI, deep learning, or NLP specialist? As a discipline, data science is maturing. Organizations have increasingly recognized the value of data science to their business, entire companies are based on AI products leveraging the power of deep learning for image recognition (e.g., Clarifai) or offering natural language generation solutions (e.g., Narrative Science). With success comes a greater recognition, appreciation of differences, and specialization. What’s more, the sheer complexity of new, emerging algorithms requires deep expertise. 2017 will see a proliferation of data roles.
The opportunity to specialize allows data people to focus on what they are good at and enjoy, great. But there will be growing pains. It takes time to understand the meaning of new job titles; companies will be advertising data science roles when they want machine learning engineers and vice versa. Hype combined with a fierce battle over talent will lead to an overabundance of “trendy” roles blurring useful differences. As a community, we will have to clarify what the new roles mean (and we’ll have to hold ourselves accountable when hiring hits a rough patch).
We will have to figure out processes for data scientists, machine learning engineers, and deep learning/AI/NLP specialists to work together productively which will affect adjacent roles. Andrew Ng, Chief Scientist at Baidu, argues for the new role of AI Product Manager who sets expectations by providing data folks with the test set (i.e., the data an already trained algorithm should perform well on). We may need transitional roles like the Chief AI Officer to guide companies in recognizing and leveraging the power of emerging algorithms.
2017 will be an exciting year for teams to experiment, but there will be battle scars.
(5) What Are You Doing To My Data?
By developing models to guide law enforcement, models to predict recidivism, models to predict job performance based on Facebook profiles, data scientists are playing high stakes games with other people’s lives. Models make mistakes; a perfectly qualified and capable candidates may not get her dream job. Data is biased; word embeddings (mentioned above) encode the meaning of words through the context in which they are used, allow simple analogies, and, trained on Google News articles, reveal gender stereotypes—”man is to programmer as woman is to homemaker”. Faulty reward functions can cause agents to go haywire. Models are powerful tools. Caution.
In 2016, Cathy O’Neill published Weapons of Math Destruction on the potential harm of algorithms which got significant attention (e.g., Scientific American, NPR). FATML, a conference on Fairness, Accountability, and Transparency in Machine Learning had record attendance. The EU issued new regulation including “the right to be forgotten”, giving individuals control over their data, and restricts the use of automated, individual decision-making especially if decisions the algorithm makes cannot be explained to the individual. Problematically, automated, individual decision-making is what neural networks do and their inner workings are hard to explain.
2017 will see companies grappling with the consequences of this “right to an explanation” which Oxford researchers have started to explore. In 2017, we may come to a refined understanding of what we mean when we say: “a model is interpretable”. Human decisions are interpretable in some sense, we can provide explanations for our decisions, but not others, we do not (yet) understand the brain dynamics underlying (complex) decisions. We will make progress on algorithms that help us understand model behavior and exercise the much needed caution when we build predictive models areas like healthcare, education, and law enforcement.
In 2017, let’s commit to responsible data science and machine learning.
– Friederike
Many thanks to Jeremy Karnowski for helpful comments.
