Written by Dave Nyczepir Feb 19, 2020 | FEDSCOOP

National Oceanic and Atmospheric Administration scientists are preparing to use machine learning (ML) to more easily monitor threatened ice seal populations in Alaska between April and May.

Ice flows are critical to seal life cycles but are melting due to climate change which has hit the Arctic and sub-Arctic regions hardest. So scientists are trying to track species population distributions.

But surveying millions of aerial photographs of sea ice a year for ice seals takes months. And the data is outdated by the time statisticians analyze it and share it with the NOAA assistant regional administrator for protected resources in Juneau, according to aMicrosoft blog post.

NOAAs Juneau office oversees conservation and recovery programs for marine mammals statewide and can instruct other agencies to limit permits for activities that might hurt species feeding or breeding. The faster NOAA processes scientific data, the faster it can implement environmental sustainability policies.

The amazing thing is how consistent these problems are from scientist to scientist, Dan Morris, principal scientist and program director of MicrosoftAI for Earth, told FedScoop.

To speed up monitoring from months to mere hours, NOAAs Marine Mammal Laboratory partnered with AI for Earth in the summer of 2018 to develop ML models recognizing seals in real-time aerial photos.

The models were trained during a one-week hackathon using 20 terabytes of historical survey data in the cloud.

In 2007, the first NOAA survey done by helicopter captured about 90,000 images that took months to analyze and find 200 seals. The challenge isthe seals are solitary, and aircraft cant fly so low as to spook them. But still, scientists need images to capture the difference between threatened bearded and ringed seals and unthreatened spotted and ribbon seals.

Alaskas rainy, cloudy climate has led scientists to adopt thermal and color cameras, but dirty ice and reflections continue to interfere. A 2016 survey of 1 million sets of images took three scientists six months to identify about 316,000 seal hotspots.

Microsofts ML, on the other hand, can distinguish seals from rocks and, coupled with improved cameras on a NOAA turboprop airplane, will be used in flyovers of the Beaufort Sea this spring.

NOAA released a finalized Artificial Intelligence Strategy on Tuesday aimed at reducing the cost of data processing and incorporating AI into scientific technologies and services addressing mission priorities.

Theyre a very mature organization in terms of thinking about incorporating AI into remote processing of their data, Morris said.

The camera systems on NOAA planes are also quite sophisticated because the agencys forward-thinking ecologists are assembling the best hardware, software and expertise for their biodiversity surveys, he added.

While the technical integration of AI for Earths models with the software systems on NOAAs planes has taken a year to perfect, another agency project was able to apply a similar algorithm more quickly.

The Cook Inlets endangered beluga whale population numbered 279 last year down from about 1,000three decades ago.

Belugas increasingly rely on echolocation to communicate with sediment from melting glaciers dirtying the water they live in. But the noise from an increasing number of cargo ships and military and commercial flights can disorient the whales. Calves can get lost if they cant hear their mothers clicks and whistles, and adults cant catch prey or identify predators.

NOAA is using ML tools to distinguish a whales whistle from man-made noises and identify areas where theres dangerous overlap, such as where belugas feed and breed. The agency can then limit construction or transportation during those periods, according to the blog post.

Previously, the projects 15 mics recorded sounds for six months along the seafloor, scientists collected the data, and then they spent the remainder of the year classifying noises to determine how the belugas spent their time.

AI for Earths algorithms matched scientists previously classified logs 99 percent of the time last fall and have been since introduced into the field.

The ML was implemented faster than the seal projects because the software runs offline at a lab in Seattle, so integration was easier, Morris said.

NOAA intends to employ ML in additional biodiversity surveys. And AI for Earth plans to announce more environmental sustainability projects in the acoustic space in the coming weeks, Morris added, thoughhe declined to name partners.

More here:
Machine learning is making NOAA's efforts to save ice seals and belugas faster - FedScoop

TAMPA, Fla. & SEATTLE--(BUSINESS WIRE)--Syniverse, the worlds most connected company, and RealNetworks, a leader in digital media software and services, today announced they have incorporated sophisticated machine learning (ML) features into their integrated offering that gives carriers visibility and control over mobile messaging traffic. By integrating RealNetworks Kontxt application-to-person (A2P) message categorization capabilities into Syniverse Messaging Clarity, mobile network operators (MNOs), internet service providers (ISPs), and messaging aggregators can identify and block spam, phishing, and malicious messages by prioritizing legitimate A2P traffic, better monetizing their service.

Syniverse Messaging Clarity, the first end-to-end messaging visibility solution, utilizes the best-in-class grey route firewall, and clearing and settlement tools to maximize messaging revenue streams, better control spam traffic, and closely partner with enterprises. The solution analyzes the delivery of messages before categorizing them into specific groupings, including messages being sent from one person to another person (P2P), A2P messages, or outright spam. Through its existing clearing and settlement capabilities, Messaging Clarity can transform upcoming technologies like Rich Communication Services (RCS) and chatbots into revenue-generating products and services without the clutter and cost of spam or fraud.

The foundational Kontxt technology adds natural language processing and deep learning techniques to Messaging Clarity to continually update and improve its understanding of messages and clarification. This new feature adds to Messaging Claritys ability to identify, categorize, and ascribe a monetary value to the immense volume and complexity of messages that are delivered through text messaging, chatbots, and other channels.

The Syniverse and RealNetworks Kontxt message classification provides companies the ability to ensure that urgent messages, like one-time passwords, are sent at a premium rate compared with lower-priority notifications, such as promotional offers. The Syniverse Messaging Clarity solution also helps eliminate instances of extreme message spam phishing (smishing). This type of attack recently occurred with a global shipping company when spam texts were sent to consumers with the request to click a link to receive an update on a package delivery for a phantom order.

CLICK TO TWEET: Block #spam and categorize & prioritize #textmessages with @Syniverse & @RealNetworks #Kontxt. #MNO #ISPs #Messaging #MachineLearning #AI http://bit.ly/2HalZkv

Supporting Quotes

Syniverse offers companies the capability to use machine learning technologies to gain insight into what traffic is flowing through their networks, while simultaneously ensuring consumer privacy and keeping the actual contents of the messages hidden. The Syniverse Messaging Clarity solution can generate statistics examining the type of traffic sent and whether it deviates from the senders traffic pattern. From there, the technology analyzes if the message is a valid one or spam and blocks the spam.

The self-learning Kontxt algorithms within the Syniverse Messaging Clarity solution allow its threat-assessment techniques to evolve with changes in message traffic. Our analytics also verify that sent messages conform to network standards pertaining to spam and fraud. By deploying Messaging Clarity, MNOs and ISPs can help ensure their compliance with local regulations across the world, including the U.S. Telephone Consumer Protection Act, while also avoiding potential costs associated with violations. And, ultimately, the consumer -- who is the recipient of more appropriate text messages and less spam -- wins as well, as our Kontxt technology within the Messaging Clarity solution works to enhance customer trust and improve the overall customer experience.

Digital Assets

Supporting Resources

About Syniverse

As the worlds most connected company, Syniverse helps mobile operators and businesses manage and secure their mobile and network communications, driving better engagements and business outcomes. For more than 30 years, Syniverse has been the trusted spine of mobile communications by delivering the industry-leading innovations in software and services that now connect more than 7 billion devices globally and process over $35 billion in mobile transactions each year. Syniverse is headquartered in Tampa, Florida, with global offices in Asia Pacific, Africa, Europe, Latin America and the Middle East.

About RealNetworks

Building on a legacy of digital media expertise and innovation, RealNetworks has created a new generation of products that employ best-in-class artificial intelligence and machine learning to enhance and secure our daily lives. Kontxt (www.kontxt.com) is the foremost platform for categorizing A2P messages to help mobile carriers build customer loyalty and drive new revenue through text message classification and antispam. SAFR (www.safr.com) is the worlds premier facial recognition platform for live video. Leading in real world performance and accuracy as tested by NIST, SAFR enables new applications for security, convenience, and analytics. For information about our other products, visit http://www.realnetworks.com.

RealNetworks, Kontxt, SAFR and the companys respective logos are trademarks, registered trademarks, or service marks of RealNetworks, Inc. Other products and company names mentioned are the trademarks of their respective owners.

Results shown from NIST do not constitute an endorsement of any particular system, product, service, or company by NIST: https://www.nist.gov/programs-projects/face-recognition-vendor-test-frvt-ongoing.

Go here to see the original:
Syniverse and RealNetworks Collaboration Brings Kontxt-Based Machine Learning Analytics to Block Spam and Phishing Text Messages - Business Wire

At HIMSS20 next month, two machine learning experts will show how machine learning algorithms are evolving to handle complex physiological data and drive more detailed clinical insights.

During surgery and other critical care procedures, continuous monitoring of blood pressure to detect and avoid the onset of arterial hypotension is crucial. New machine learning technology developed by Edwards Lifesciences has proven to be an effective means of doing this.

In the prodromal stage of hemodynamic instability, which is characterized by subtle, complex changes in different physiologic variables unique dynamic arterial waveform "signatures" are formed, which require machine learning and complex feature extraction techniques to be utilized.

Feras Hatib, director of research and development for algorithms and signal processing at Edwards Lifesciences, explained his team developed a technology that could predict, in real-time and continuously, upcoming hypertension in acute-care patients, using an arterial pressure waveforms.

We used an arterial pressure signal to create hemodynamic features from that waveform, and we try to assess the state of the patient by analyzing those signals, said Hatib, who is scheduled to speak about his work at HIMSS20.

His teams success offers real-world evidence as to how advanced analytics can be used to inform clinical practice by training and validating machine learning algorithms using complex physiological data.

Machine learning approaches were applied to arterial waveforms to develop an algorithm that observes subtle signs to predict hypotension episodes.

In addition, real-world evidence and advanced data analytics were leveraged to quantify the association between hypotension exposure duration for various thresholds and critically ill sepsis patient morbidity and mortality outcomes.

"This technology has been in Europe for at least three years, and it has been used on thousands of patients, and has been available in the US for about a year now," he noted.

Hatib noted similar machine learning models could provide physicians and specialists with information that will help prevent re-admissions or other treatment options, or help prevent things like delirium current areas of active development.

"In addition to blood pressure, machine learning could find a great use in the ICU, in predicting sepsis, which is critical for patient survival," he noted. "Being able to process that data in the ICU or in the emergency department, that would be a critical area to use these machine learning analytics models."

Hatib pointed out the way in which data is annotated in his case, defining what is hypertension and what is not is essential in building the machine learning model.

"The way you label the data, and what data you include in the training is critical," he said. "Even if you have thousands of patients and include the wrong data, that isnt going to help its a little bit of an art to finding the right data to put into the model."

On the clinical side, its important to tell the clinician what the issue is in this case what is causing hypertension.

"You need to provide to them the reasons that could be causing the hypertension this is why we complimented the technology with a secondary screen telling the clinician what is physiologically is causing hypertension," he explained. "Helping them decide what do to about it was a critical factor."

Hatib said in the future machine learning will be everywhere, because scientists and universities across the globe are hard at work developing machine learning models to predict clinical conditions.

"The next big step I see is going toward using this ML techniques where the machine takes care of the patient and the clinician is only an observer," he said.

Feras Hatib, along with Sibyl Munson of Boston Strategic Partners, will share some machine learning best practices during his HIMSS20 in a session, "Building a Machine Learning Model to Drive Clinical Insights." It's scheduled for Wednesday, March 11, from 8:30-9:30 a.m. in room W304A.

The rest is here:
Machine learning and clinical insights: building the best model - Healthcare IT News

Pluto7 is a services and solutions company focused on accelerating business transformation. As a Google Cloud Premier Partner, we service the retail, manufacturing, healthcare, and hi-tech industries.

Pluto7 just achieved the Google Cloud Machine Learning Specialization for combining business consultancy and unique machine learning solutions built on Google Cloud.

With Pluto7 comes unique capabilities for machine learning, artificial intelligence, and analytics. Brought to you by a company that contains some of the finest minds in data science, able to draw on its surroundings in the very heart of Silicon Valley, California.

Businesses are looking for practical solutions to real-world challenges. And by that, we do not just mean providing the tech and leaving you to stitch it all together. Instead, Pluto7s approach is to apply innovation to your desired outcome, alongside the experience needed to make it all happen. This is where their range of consultancy services comes into play. These are designed to create an interconnected tech stack and to champion data empowerment through ML/AI.

Pluto7s services and solutions allow businesses to speed up and scale-out sophisticated machine learning models. They have successfully guided many businesses through the digital transformation process by leveraging the power of artificial intelligence, analytics, and IoT solutions.

What does this mean for a partner to be specialized?

When you see a Google Cloud partner with a Specialization, it indicates proficiency and experience with Google Cloud. Pluto7 is recognized by Google Cloud as a machine learning specialist with deep technical capabilities. The organizations that receive this distinction, demonstrates their ability to lead a customer through the entire AI journey. Pluto7 designs, builds, migrates, tests, and operates industry-specific solutions for their customers.

Pluto7 has a plethora of previous experience in deploying accelerated solutions and custom applications in machine learning and AI. The many proven success stories from industry leaders like ABinBev, DxTerity, L-Nutra, CDD, USC, UNM are publically available on their website. These customers have leveraged Pluto7 and Google Cloud technology to see tangible and transformative results.

On top of all this, Pluto7 has a business plan that aligns with the Specialization. Because of their design, build, and implementation methodologies they are able to successfully drive innovation, accelerate business transformation, and boost human creativity.

ML Services and Solutions

Pluto7 has created Industry-specific use cases for marketing, sales, and supply chains and integrated these to deliver a game-changing customer experience. These capabilities are brought to life through their partnership with Google Cloud, one of the most innovative platforms for AI and ML out there. The following solution suites are created to solve some of the most difficult problems through a combination of innovative technology and deep industry expertise.

Demand ML - Increase efficiency and lower costs

Pluto7 helps supply chain leaders manage unpredictable fluctuations. These solutions allow businesses to achieve demand forecast accuracy of more than 90%, manage complex and unpredictable fluctuations while delivering the right product at the right time -- all using AI to predict and recommend based on real-time data at scale.

Preventive Maintenance - Improve quality, production and reduce associated costs

Pluto7 improves the production efficiency of production plants from 45-80% to reduce downtime and maintain quality. They leverage machine learning and predictive analytics to determine the remaining value of assets and accurately determine when a manufacturing plant, machine, component or part is likely to fail, and thus needs to be replaced.

Marketing ML - Increase marketing ROI

Pluto7s marketing solutions improve click-through rates and predict traffic rates accurately. Pluto7 can help you analyze marketing data in real-time to transform prospect and customer engagement with hyper-personalization. Businesses are able to leverage machine learning for better customer segmentation, campaign targeting, and content optimization.

Contact Pluto7

If you would like to begin your AI journey, Pluto7 recommends starting with a discovery workshop. This workshop is co-driven by Pluto7 and Google Cloud to understand business pain points and set up a strategy to begin solving. Visit the website at http://www.pluto7.com and contact us to get started today!

View source version on businesswire.com: https://www.businesswire.com/news/home/20200219005054/en/

Contacts

Sierra ShepardGlobal Marketing Teammarketing@pluto7.com

Excerpt from:
Pluto7, a Google Cloud Premier Partner, Achieved the Machine Learning Specialization and is Recognized by Google Cloud as a Machine Learning...

A few weeks ago I wrote a piece here about Moogsoft which has been making waves in the service assurance space by applying artificial intelligence and machine learning to the arcane task of keeping on keeping critical IT up and running and lessening the business impact of service interruptions. Its a hot area for startups and Ive since gotten article pitches from several other AIops firms at varying levels of development.

The most intriguing of these is a company called Grok which was formed by a partnership between Numenta, a pioneering AI research firm co-founded by Jeff Hawkins and Donna Dubinsky, who are famous for having started two classic mobile computing companies, Palm and Handspring, and Avik Partners. Avik is a company formed by brothers Casey and Josh Kindiger, two veteran entrepreneurs who have successfully started and grown multiple technology companies in service assurance and automation over the past two decadesmost recently Resolve Systems.

Josh Kindiger told me in a telephone interview how the partnership came about:

Numenta is primarily a research entity started by Jeff and Donna about 15 years ago to support Jeffs ideas about the intersection of neuroscience and data science. About five years ago, they developed an algorithm called HTM and a product called Grok for AWS which monitors servers on a network for anomalies. They werent interested in developing a company around it but we came along and saw a way to link our deep domain experience in the service management and automation areas with their technology. So, we licensed the name and the technology and built part of our Grok AIOps platform around it.

Jeff Hawkins has spent most of his post-Palm and Handspring years trying to figure out how the human brain works and then reverse engineering that knowledge into structures that machines can replicate. His model or theory, called hierarchical temporal memory (HTM), was originally described in his 2004 book On Intelligence written with Sandra Blakeslee. HTM is based on neuroscience and the physiology and interaction of pyramidal neurons in the neocortex of the mammalian (in particular, human) brain. For a little light reading, I recommend a peer-reviewed paper called A Framework for Intelligence and Cortical Function Based on Grid Cells in the Neocortex.

Grok AIOps also uses traditional machine learning, alongside HTM. Said Kindiger:

When I came in, the focus was purely on anomaly detection and I immediately engaged with a lot of my old customers--large fortune 500 companies, very large service providers and quickly found out that while anomaly detection was extremely important, that first signal wasn't going to be enough. So, we transformed Grok into a platform. And essentially what we do is we apply the correct algorithm, whether it's HTM or something else, to the proper stream events, logs and performance metrics. Grok can enable predictive, self-healing operations within minutes.

The Grok AIOps platform uses multiple layers of intelligence to identify issues and support their resolution:

Anomaly detection

The HTM algorithm has proven exceptionally good at detecting and predicting anomalies and reducing noise, often up to 90%, by providing the critical context needed to identify incidents before they happen. It can detect anomalies in signals beyond low and high thresholds, such as signal frequency changes that reflect changes in the behavior of the underlying systems. Said Kindiger:

We believe HTM is the leading anomaly detection engine in the market. In fact, it has consistently been the best performing anomaly detection algorithm in the industry resulting in less noise, less false positives and more accurate detection. It is not only best at detecting an anomaly with the smallest amount of noise but it also scales, which is the biggest challenge.

Anomaly clustering

To help reduce noise, Grok clusters anomalies that belong together through the same event or cause.

Event and log clustering

Grok ingests all the events and logs from the integrated monitors and then applies to it to event and log clustering algorithms, including pattern recognition and dynamic time warping which also reduce noise.

IT operations have become almost impossible for humans alone to manage. Many companies struggle to meet the high demand due to increased cloud complexity. Distributed apps make it difficult to track where problems occur during an IT incident. Every minute of downtime directly impacts the bottom line.

In this environment, the relatively new solution to reduce this burden of IT management, dubbed AIOps, looks like a much needed lifeline to stay afloat. AIOps translates to "Algorithmic IT Operations" and its premise is that algorithms, not humans or traditional statistics, will help to make smarter IT decisions and help ensure application efficiency. AIOps platforms reduce the need for human intervention by using ML to set alerts and automation to resolve issues. Over time, AIOps platforms can learn patterns of behavior within distributed cloud systems and predict disasters before they happen.

Grok detects latent issues with cloud apps and services and triggers automations to troubleshoot these problems before requiring further human intervention. Its technology is solid, its owners have lots of experience in the service assurance and automation spaces, and who can resist the story of the first commercial use of an algorithm modeled on the human brain.

Read more:
Grok combines Machine Learning and the Human Brain to build smarter AIOps - Diginomica

The report on the global machine learning as a service market provides qualitative and quantitative analysis for the period from 2016 to 2024. The report predicts the global machine learning as a service market to grow with a CAGR of 38.

New York, Feb. 20, 2020 (GLOBE NEWSWIRE) -- Reportlinker.com announces the release of the report "Machine Learning as a Service Market: Global Industry Analysis, Trends, Market Size, and Forecasts up to 2024" - https://www.reportlinker.com/p05751673/?utm_source=GNW 5% over the forecast period from 2018-2024. The study on machine learning as a service market covers the analysis of the leading geographies such as North America, Europe, Asia-Pacific, and RoW for the period of 2016 to 2024.

The report on machine learning as a service market is a comprehensive study and presentation of drivers, restraints, opportunities, demand factors, market size, forecasts, and trends in the global machine learning as a service market over the period of 2016 to 2024. Moreover, the report is a collective presentation of primary and secondary research findings.

Porters five forces model in the report provides insights into the competitive rivalry, supplier and buyer positions in the market and opportunities for the new entrants in the global machine learning as a service market over the period of 2016 to 2024. Further, IGR- Growth Matrix gave in the report brings an insight into the investment areas that existing or new market players can consider.

Report Findings1) Drivers Increasing use in cloud technologies Provides statistical analysis along with reduce time and cost Growing adoption of cloud based systems2) Restraints Less skilled personnel3) Opportunities Technological advancement

Research Methodology

A) Primary ResearchOur primary research involves extensive interviews and analysis of the opinions provided by the primary respondents. The primary research starts with identifying and approaching the primary respondents, the primary respondents are approached include1. Key Opinion Leaders associated with Infinium Global Research2. Internal and External subject matter experts3. Professionals and participants from the industry

Our primary research respondents typically include1. Executives working with leading companies in the market under review2. Product/brand/marketing managers3. CXO level executives4. Regional/zonal/ country managers5. Vice President level executives.

B) Secondary ResearchSecondary research involves extensive exploring through the secondary sources of information available in both the public domain and paid sources. At Infinium Global Research, each research study is based on over 500 hours of secondary research accompanied by primary research. The information obtained through the secondary sources is validated through the crosscheck on various data sources.

The secondary sources of the data typically include1. Company reports and publications2. Government/institutional publications3. Trade and associations journals4. Databases such as WTO, OECD, World Bank, and among others.5. Websites and publications by research agencies

Segment CoveredThe global machine learning as a service market is segmented on the basis of component, application, and end user.

The Global Machine Learning As a Service Market by Component Software Services

The Global Machine Learning As a Service Market by Application Marketing & Advertising Fraud Detection & Risk Management Predictive Analytics Augmented & Virtual Reality Security & Surveillance Others

The Global Machine Learning As a Service Market by End User Retail Manufacturing BFSI Healthcare & Life Sciences Telecom Others

Company Profiles IBM PREDICTRON LABS H2O.ai. Google LLC Crunchbase Inc. Microsoft Yottamine Analytics, LLC Fair Isaac Corporation. BigML, Inc. Amazon Web Services, Inc.

What does this report deliver?1. Comprehensive analysis of the global as well as regional markets of the machine learning as a service market.2. Complete coverage of all the segments in the machine learning as a service market to analyze the trends, developments in the global market and forecast of market size up to 2024.3. Comprehensive analysis of the companies operating in the global machine learning as a service market. The company profile includes analysis of product portfolio, revenue, SWOT analysis and latest developments of the company.4. IGR- Growth Matrix presents an analysis of the product segments and geographies that market players should focus to invest, consolidate, expand and/or diversify.Read the full report: https://www.reportlinker.com/p05751673/?utm_source=GNW

About ReportlinkerReportLinker is an award-winning market research solution. Reportlinker finds and organizes the latest industry data so you get all the market research you need - instantly, in one place.

__________________________

Story continues

Clare: clare@reportlinker.comUS: (339)-368-6001Intl: +1 339-368-6001

Read the original here:
Global machine learning as a service market is expected to grow with a CAGR of 38.5% over the forecast period from 2018-2024 - Yahoo Finance

It is much easier to apologize than it is to get permission.

jamesnoellert.com

The hacking culture has been the lifeblood of software engineering long before the move fast and break things mantra became ubiquitous of tech startups [1, 2]. Computer industry leaders from Chris Lattner [3] to Bill Gates recount breaking and reassembling radios and other gadgets in their youth, ultimately being drawn to computers for their hackability. Silicon Valley itself may have never become the worlds innovation hotbed if it were not for the hacker dojo started by Gordon French and Fred Moore, The Homebrew Club.

Computer programmers still strive to move fast and iterate things, developing and deploying reliable, robust software by following industry proven processes such as test-driven development and the Agile methodology. In a perfect world, programmers could follow these practices to the letter and ship pristine software. Yet time is money. Aggressive, business-driven deadlines pass before coders can properly finish developing software ahead of releases. Add to this the modern best practices of rapid-releases and hot-fixing (or updating features on the fly [4]), the bar for deployable software is even lower. A company like Apple even prides itself by releasing phone hardware with missing software features: the Deep Fusion image processing was part of an iOS update months after the newest iPhone was released [5].

Software delivery becoming faster is a sign of progress; software is still eating the world [6]. But its also subject to abuse: Rapid software processes are used to ship fixes and complete new features, but are also used to ship incomplete software that will be fixed later. Tesla has emerged as a poster child with over the air updates that can improve driving performance and battery capacity, or hinder them by mistake [7]. Naive consumers laud Tesla for the tech-savvy, software-first approach theyre bringing to the old-school automobile industry. Yet industry professionals criticize Tesla for their recklessness: A/B testing [8] an 1800kg vehicle on the road is slightly riskier than experimenting with a new feature on Facebook.

Add Tesla Autopilot and machine learning algorithms into the mix, and this becomes significantly more problematic. Machine learning systems are by definition probabilistic and stochastic predicting, reacting, and learning in a live environment not to mention riddled with corner cases to test and vulnerabilities to unforeseen scenarios.

Massive progress in software systems has enabled engineers to move fast and iterate, for better or for worse. Now with massive progress in machine learning systems (or Software 2.0 [9]), its seamless for engineers to build and deploy decision-making systems that involve humans, machines, and the environment.

A current danger is that the toolset of the engineer is being made widely available but the theoretical guarantees and the evolution of the right processes are not yet being deployed. So while deep learning has the appearance of an engineering profession it is missing some of the theoretical checks and practitioners run the risk of falling flat upon their faces.

In his recent book Reboot AI [10], Gary Marcus draws a thought provoking analogy between deep learning and pharmacology: Deep learning models are more like drugs than traditional software systems. Biological systems are so complex it is rare for the actions of medicine to be completely understood and predictable. Theories of how drugs work can be vague, and actionable results come from experimentation. While traditional software systems are deterministic and debuggable (and thus robust), drugs and deep learning models are developed via experimentation and deployed without fundamental understanding and guarantees. Too often the AI research process is first experiment, then justify results. It should be hypothesis-driven, with scientific rigor and thorough testing processes.

What were missing is an engineering discipline with principles of analysis and design.

Before there was civil engineering, there were buildings that fell to the ground in unforeseen ways. Without proven engineering practices for deep learning (and machine learning at large), we run the same risk.

Taking this to the extreme is not advised either. Consider the shift in spacecraft engineering the last decade: Operational efficiencies and the move fast culture has been essential to the success of SpaceX and other startups such as Astrobotic, Rocket Lab, Capella, and Planet.NASA cannot keep up with the pace of innovation rather, they collaborate with and support the space startup ecosystem. Nonetheless, machine learning engineers can learn a thing or two from an organization that has an incredible track record of deploying novel tech in massive coordination with human lives at stake.

Grace Hopper advocated for moving fast: That brings me to the most important piece of advice that I can give to all of you: if you've got a good idea, and it's a contribution, I want you to go ahead and DO IT. It is much easier to apologize than it is to get permission. Her motivations and intent hopefully have not been lost on engineers and scientists.

[1] Facebook Cofounder Mark Zuckerberg's "prime directive to his developers and team", from a 2009 interview with Business Insider, "Mark Zuckerberg On Innovation".

[2] xkcd

[3] Chris Lattner is the inventor of LLVM and Swift. Recently on the AI podcast, he and Lex Fridman had a phenomenal discussion:

[4] Hotfix: A software patch that is applied to a "hot" system; i.e., a fix to a deployed system already in use. These are typically issues that cannot wait for the next release cycle, so a hotfix is made quickly and outside normal development and testing processes.

[5]

[6]

[7]

[8] A/B testing is an experimental processes to compare two or more variants of a product, intervention, etc. This is very common in software products when considering e.g. colors of a button in an app.

[9] Software 2.0 was coined by renowned AI research engineer Andrej Karpathy, who is now the Director of AI at Tesla.

[10]

[11]

Visit link:
Machine Learning Is No Place To Move Fast And Break Things - Forbes