Category: Big-Data

Luminar, a leading maker of laser lidar for self-driving and semi-automated vehicles, is accused of passing off a next-generation chip design created by a rival as its own technology after showing an image of the processor at a recent investor conference and in materials on its website. Lidwave, the Israeli startup making the claim, says it plans to take legal action over the matter.

The image, identified as a photonic integrated circuit by Luminar in its Feb. 28 conference and webcast with no reference to Lidwave, looks identical to a chip on Lidwave’s website that is its core technology. The Jerusalem-based company sent a cease-and-desist letter to Luminar on March 14 asking it to remove the image. It also notified the Securities and Exchange Commission of Luminar’s “misuse of its product image to falsely promote its abilities and securities to investors.”

Lidwave said it hadn’t received a response from the company as of 3 pm New York time on Friday. The SEC did not immediately respond to a request for comment on the matter.

“Integrated photonics allows lidar to become as scalable and profitable as anybody has imagined … But surprisingly, the picture Luminar presented of a photonic integrated circuit as their solution is not their solution, but basically the exact image of our lidar,” Lidwave CEO Yehuda Vidal told Forbes. “We will continue with a lawsuit if needed because it’s a very huge problem for us. Some of our customers are very confused about what we do versus what they do.”

The image appeared during a presentation by Mike McCauliffe, head of Luminar’s semiconductor team. The full Luminar Day webcast on YouTube has nearly 750,000 views. “The presenter couldn’t confirm the original image source, but we are replacing it,” company spokesman Milin Mehta told Forbes.

The company did not respond to Lidwave’s other claims. The chip image remains in Luminar’s presentation and YouTube video as of the time of publication.

Lidar’s ability to create detailed, 3D maps of a vehicle’s surroundings using lasers has made it a core technology in the race to perfect self-driving cars and trucks. But it’s also relatively expensive, with individual units costing thousands of dollars each. Luminar, which has high-volume supply agreements with automakers including Mercedes-Benz, Volvo, Nissan, Polestar, China’s SAIC and truckmaker Daimler, is poised to be the biggest supplier of the technology by driving down the cost dramatically in the years to come.

Over the past five years, Orlando-based Luminar, created by optics prodigy (and Forbes 30 Under 30) alum Austin Russell moved from being one of dozens of lidar startups vying to challenge Velodyne, the first company to commercialize the technology for autonomous vehicles, to the de facto industry leader, in terms of the number of vehicles that will be using its sensors. Russell told Forbes last month that he intends to have Luminar lidar in “millions of vehicles on the road within a few years.”

Processing vast amounts of data collected by laser sensors requires ever more powerful chips that must also be cheaper and easier to produce. Lidwave’s Vidal said that is exactly what his company, formed in 2019 and hoping to begin commercial deliveries in 2024, is working to perfect.

“The main problem holding back wide-scale adoption of lidar is its cost resulting from extremely complex micro-precision assembly. Today there is no lidar producer that’s conclusively solved the economy-of-scale (production) challenge,” he said. “A photonic integrated chip platform is basically the Holy Grail for lidar.”

Ecommerce professionals face a tsunami-sized flood of data to make decisions. In today’s intensely competitive digital world, the decisions made by ecommerce professionals can significantly impact the growth of their business and their capacity to outperform competitors. As we all know, data-driven decisions are the gold standard for ensuring success, but while data is here to help, it is actually getting in the way, and right now the crux of the issue lies in the sheer volume of data that is readily available.

The sheer magnitude of data that ecommerce professionals must sift through in order to make well-informed decisions is more than the number of books in the entire British library. They rely on web analytics to understand website performance, customer data for insights into peoples preferences and needs, sales data for evaluating product performance, and then social media analytics to track customer engagement and any commercial traction and best in class content. Additionally, market research provides context on industry trends, whilst email marketing, product data, and advertising data offer valuable information for refining marketing strategies, managing inventory, and optimising advertising efforts. Ecommerce leaders often have lengthy meetings discussing reports, sometimes up to 20 pages long, that inevitably obscure crucial insights and trends that are essential for decision making. Frankly speaking, this uncontrollable flood of data would overwhelm any human being.

Humans have a limited capacity to process data. Research has shown that when inundated with an unmanageable deluge of information, we tend to feel stressed. Our cognitive systems default to habits that hinder our decision-making abilities, preventing us from using the most reliable evidence at our disposal, ultimately leading to compromised performance and missed opportunities.

There’s hope, though. Based on interviews with ecommerce leaders like Clare Kelly from Sass & Belle, Mark Morley from Prep Kitchen, Ryan Clarke from StatSports UK and Gareth Johnson from Overclockers, it was clear that data is a cornerstone for ecommerce leaders to succeed, but you need to put in some effort to get it right to drive value. Until we get platforms that automatically sort and prioritise the data that we need to pay attention to, there are manual things that you can do to address the problem.

1. Prioritise relevant information and insights aligned with your business objectives. Avoid drowning in a sea of data by working with your team to pinpoint metrics crucial for success, then concentrate on gathering and analysing only those key metrics.

2. Present data in a clear and concise format. Data reports must emphasise key insights, trends, and actionable recommendations for improvement. Begin to use data stories to comprehend complex datasets better and then swiftly identify important patterns.

3. Utilise relevant benchmarks: Gauge your performance against industry standards and best practices to identify improvement areas and make data-driven decisions.

4. Leverage predictive analytics to stay ahead of the competition by anticipating future trends, seizing opportunities, and mitigating threats.

5. Build a suitable cadence for data to stay updated on trends and insights for informed decision-making.

Developing this can take time, but you’ll see the benefits when you do.

Analysing your online store(s) will become effortless, enabling you to assess the impact of particular initiatives, campaigns and tactics and the efficacy of the agencies you have hired to help you grow. You will be able to identify areas where substantial revenues are being missed and formulate strategies to outperform your competitors.

The success of an ecommerce business depends heavily on this effective use of data. While too much data can hinder progress, focusing on the steps outlined above will empower you to make data-informed decisions that will drive growth, improve performance and help digital businesses reach the growth targets they have set out.

Smart leaders will address this problem sooner rather than later.

Search is changing. Humans (and in some cases software machine ‘bots’ working inside automation engines) are searching for content, data and information in new ways. Google is still there i.e. the core principles defining web search are comparatively constant in relation to the way search is changing for enterprise use cases.

Because we are searching for information in new ways, the way data analytics works has to change too. With many users looking for information sources via mobile devices and/or by using speech, the way we frame and phrase search in our heads has evolved. We have moved towards using an increased level of idiomatically nuanced and more informally phrased language that requires contextual interpretation and additional processing to work with.

Thankfully, Artificial Intelligence (AI) and Natural Langauge Understanding (NLU) technologies have kept pace with much of this change. So-called Large Language Models (LLM) technologies have brought us GPT-3 (as in ChatGPT) and its ability to produce human-like text.

What is a large language model?

As explained here a large language model isn’t just large. The size of its datasets do define the way we regard this type of data model, but it is the neural interwoven nature of semantic context between words and phrases that enables us to create a more human level of understanding. Given that the word “set” refers to a type of yoghurt, a group of numbers, a starting call for a running race, something the sun does, a collection of dishes, the action of putting an object down, a television studio and a place where badgers live – you can see why we need large language models to define the relationships between one word and another.

Where these technologies, nuances of language and mathematical models all now meet is in data analytics designed to enable us to ‘do stuff’ in business and search for things.

Given the IT industry’s current preference for applying generative AI to everything from ERP to CRM and everything in between, now appears to be a logical time for AI-powered analytics company ThoughtSpot to unveil its integration with the GPT-3 large language model (LLM) from OpenAI. The company says it aims to make analytics accessible, trusted and scalable through AI with the launch of its ThoughtSpot Sage ‘search experience’ that combines LLMs (like GPT-3) with its own patented search technology.

The Mountain View-headquartered data analytics democratization specialist claims that any and every employee, regardless of their skill level, can confidently use AI-Powered analytics on even the largest most complex data to ask and answer business questions. ThoughtSpot Sage already supports GPT-3 at its current preview stage with support planned for additional LLMs.

Self-service data analytics

“With ThoughtSpot Sage, ThoughtSpot is taking analytics to the next level through advanced natural language processing (NLP) and generative AI. In doing so, data analytics teams are able to exponentially increase their impact across an organization by freeing themselves from tediously managing report requests as business users self-serve insights,” notes the company, in a product statement.

The natural language search technology here means users can search existing analytical content, or automatically create new charts and visualizations based on natural language search. ThoughtSpot Sage then takes this further, using AI to create narratives explaining what insight is being surfaced, so users can understand a data visualization and act upon it.

The technology here is also aligned for use by more technical users with professional engineering qualifications. With AI-assisted data modeling that incorporates natural language, data scienticts and other software engineers can create search data models that allow AI to answer questions, while preserving the reliability and guardrails of the data model itself. Synonyms can be generated quickly using LLMs, shortening the time needed for search models by making them immediately usable to business users. Furthermore, ThoughtSpot Sage can also automatically generate descriptions, helping business users understand what data they are utilizing and improving data literacy in the process.

“As a company, we made three strategic bets about what would define the next leader in analytics. First, business users would demand personalized, useful insights from their data. Second, when data moves to the cloud, customers want flexibility and freedom to choose the best tool for their specific use case. Last, we bet that user experience would crown the next leader in analytics,” said Sumeet Arora, chief development officer, ThoughtSpot.

Arora suggests that large language models like GPT-3 have made each of these three defining factors even more important. His team calls the previous decade’s data visualization tools ‘incongruent’ with the way we are using today’s cloud services an enterprise applications in the modern age.

The trust factor

It sounds like we’ve moved on, but is this all safe?

We now set off and empower businesspeople with data analytics search functionality and allow them to point the sharp end of their citizen data science weapon anywhere they feel needs a good poke and, subsequently, they start to invoke business decisions based upon the ‘insights’ that they uncover and change the course of teams, department or an entire company’s operational plan? Not as such.

ThougtSpot and other enterprise software vendors that have built tools of this nature generally take a pretty responsible attitude towards setting controls on the way the system is used. We’ve mentioned guardrails to the data model above and that control thread extends further. With ThoughtSpot Sage, the company says that customers can take advantage of the natural language of LLMs paired with the under-the-hood capabilities. Analysts and other data professionals can quickly build reusable search data models that provide guardrails for end users, so they can safely explore with natural language.

This makes it possible to leverage the power of LLMs for analytics in an actual business environment, with real business data. Business users can then use natural language to explore all their organization’s existing analytical content to find answers that have already been created or create new insights.

It’s true that GPT-3 has ushered in a new era for search, redefining the relationship between humans and machines in the process. Search is no longer simply the means to retrieve information and we will likely see our definition of what we mean by ‘search’ even further extended when we look at technology vendors pairing semantic intelligence with vector-based search and Machine Learning (ML) advancements. But that’s another story for another day.

Search is changing, language interpretation is changing and analytics is changing… and if you have any doubts about that, do a search, the old kind.

Shou Zi Chew is making the rounds in Washington and talking to members of Congress he must answer to later this month, attempting to allay their concerns about TikTok before his closely-watched inaugural moment in the hot seat.

TikTok CEO Shou Zi Chew is meeting with lawmakers on Capitol Hill ahead of his first-ever testimony before Congress, including several representatives who will be grilling him under oath on March 23.

Chew has sought closed-door meetings with at least half a dozen members of the House Energy and Commerce Committee ahead of its hearing on TikTok’s child safety issues, handling of user data and apparent ties to China, according to two senior Democratic staffers. He has met with several, including Reps. Lori Trahan of Massachusetts, Jan Schakowsky of Illinois and Scott Peters of California.

“He’s operating from a place where no one has trust in them, and he fully recognizes that,” Trahan said in an interview with Forbes after her Wednesday meeting with Chew in Washington.

“TikTok is in a really unique position right now to take some positive steps on issues that a lot of top American companies have fallen behind, and frankly even regressed, on—and I made clear to Mr. Chew that I hope to see him move to fulfill that potential,” she added.

Trahan said her meeting focused primarily on the dangers that TikTok’s younger users are facing on the platform. At the face-to-face with Schakowsky in February, according to a source present, Chew talked about TikTok’s proposed national security agreement pending with the Committee on Foreign Investment in the U.S. TikTok has been negotiating with CFIUS since 2019 on a deal that aims to address national security concerns related to the company’s ownership by Beijing-based ByteDance.

Chew’s meetings with leaders of the congressional committee to whom he’ll soon answer come as TikTok tries to calm critics in the face of record scrutiny of the Chinese-owned video app. Recent Forbes reporting revealing that ByteDance had tracked multiple journalists who cover the company, gaining access to their user data, has raised fears across the U.S. government that such surveillance could be conducted by China on Americans more broadly. And Forbes reporting detailing how TikTok has become a useful tool for child predators, and the pervasiveness of illegal sexual abuse material of minors on the platform, has drawn scrutiny in both chambers of Congress. TikTok’s charm offensive addressing some of these issues has seen the company offering press tours of a new “Transparency and Accountability Center” in Los Angeles and inviting creators to the Hill to speak to policymakers about the benefits of the platform.

TikTok did not immediately respond to a request for comment.

Lawmakers are intensifying efforts to respond to concerns about TikTok as the Biden administration struggles to strike a CFIUS deal with the company; state attorneys general investigate the app’s alleged harms to minors; and state and federal lawmakers try to restrict or outright shut down the app in the U.S. A week after the House Foreign Affairs Committee voted to advance a Republican-led, TikTok-focused bill that would enable President Joe Biden to ban it, a dozen senators introduced broader bipartisan legislation empowering the Department of Commerce to crack down on, or even ban, communications technologies (including TikTok) built by foreign adversaries China, Iran, Russia and North Korea. The White House has endorsed that proposal, the RESTRICT Act, and separately ordered federal agencies to wipe TikTok from government employees’ devices by the end of this month.

Some lawmakers welcome courtesy meetings with CEOs before they get hauled before Congress to allow for more time to go deeper on certain issues, while others have a blanket policy to not take them. Despite the heightened anxiety around TikTok, a senior Democratic staffer present at one of Chew’s recent lawmaker chats described it as fairly typical relative to past closed-door, pre-hearing sessions with Meta CEO Mark Zuckerberg, Google CEO Sundar Pichai and former YouTube CEO Susan Wojcicki.

Trahan, for her part, said her chat with Chew this week was more substantive than past discussions she’s had with the heads of Meta, Instagram and Google. “Mr. Chew didn’t practice the art of deflection; there were frank conversations around the harms that he knows exist… how they’re mitigating risks on the platform, and where they need help from Congress and what they need to do better,” she told Forbes. “It was just a more honest exchange than any I’ve had with the other American CEOs.” The conversation also highlighted “how disappointing it is” that many of the same issues with young users remain widespread on domestic platforms that have had far more time than TikTok to fix them, she said.

Trahan added that tech CEO hearings, like the much-anticipated one with Chew on March 23, can sometimes “turn into political theater.”

“I don’t think that’s helpful… and we haven’t taken up any major legislation that would do anything meaningful in the tech space. We really need to continue to do our jobs in this regard,” she told Forbes. “We’ve got privacy legislation, we have transparency legislation, and if we just allow these hearings to be political theater, we will have done nothing to protect the American people.”

AI is all-pervasive. The penetration of Artificial Intelligence (AI) and Machine Learning (ML) is increasing the intelligence quotient in enterprise software applications in every sphere and across all industries.

The next wave of AI may very likely be characterized by its deployment into dedicated application types, with software engineering and algorithmic power aligned to serve more specifically segmented tasks – rather than it ‘just’ being data-driven smartness capable of collating and analyzing information of any general type.

Cloud-centric Customer Relationship Management (CRM) company Salesforce is following (the company would no doubt say leading) this trend with its launch of Einstein GPT, a generative AI CRM technology designed to deliver AI-created content across sales, service, marketing, commerce and IT interactions.

What GPT means

Employing the now-customary use of GPT from the deep learning human language training model of the same name (as in ChatGPT), the acronym itself stands for Generative Pretrained Transformer, with the generative aspect here being a statistical classification that denotes the nature of system behavior. Salesforce promises that Einstein GPT will transform every customer experience with generative AI.

But how does it work?

Einstein GPT infuses Salesforce’s own proprietary AI models with generative AI technology from a group of partners and real-time data from the Salesforce Data Cloud (a cloud computing service built to organize and unify data from a customer’s Salesforce use cases and third-party sources as well) to ingest and harmonize all of a company’s customer data.

Crucially then, at that point, with Einstein GPT, customers can then connect that data to advanced AI models operated by OpenAI, the US research laboratory behind ChatGPT. Customers can connect to OpenAI’s AI models out of the box (i.e. directly from this Salesforce service) or choose their own external model. They can then use natural-language prompts directly within their Salesforce CRM to generate content that continuously adapts to changing customer information and needs in real-time.

Specifically, to clarify what is happening here, Salesforce is combining OpenAI’s enterprise-grade ChatGPT technology with Salesforce’s own private AI models to deliver what it insists is both relevant and trusted AI-generated CRM content.

For example, says Salesforce, “Einstein GPT can generate personalized emails for salespeople to send to customers, generate specific responses for customer service professionals to more quickly answer customer questions, generate targeted content for marketing people to increase campaign response rates and auto-generate code.”

Einstein GPT in CRM

Created all the way back in 2014, Einstein GPT is the next generation Salesforce’s Einstein AI technology. The company measures Einstein’s processing workload and says that it is currently delivering more than 200 billion AI-powered predictions per day. By combining proprietary Einstein AI models with leading large language models, customers can use natural-language prompts on CRM data to trigger time-saving automations and create personalized AI-generated content.

In terms of the actual services now being launched, the product set breaks down as follows: Einstein GPT for Service – enables users to generate knowledge articles from past case notes and auto-generate personalized agent chat replies to (attempt to) increase customer satisfaction through personalized and expedited service interactions; Einstein GPT for Sales – auto-generate sales tasks like composing emails, scheduling meetings and preparing for the next interaction; plus additional Einstein GPT services for marketing, for Slack to deliver AI-power in Slack like smart summaries of sales opportunities and background research on accounts.

There’s also Einstein GPT for Developers to improve developer productivity with Salesforce, because software application development underpins the entire tale here, obviously.

Where will generative AI work?

It’s a burgeoning field, clearly. Not only is AI usage increasing its penetration in enterprise software, it is also widening its scope into industry- and use case-specific application models as we see here in CRM. Studies into generative AI suggest that most IT managers think this technology will improve customer service, enable their teams to work better with data and also be able to operate workflows more effectively.

Perhaps most importantly of all – because of the absolutely core and fundamental fact that we know that AI is only ever as smart as the datasets we expose it to and the data models upon which it runs – IT leaders already agree that generative AI should combine public and private data sources, as is the case here.

What would Albert Einstein himself make of all this?

He would remind us that it was he himself who said, “The only source of knowledge is experience.”

Networks are accelerating. As a general rule, almost everything the collective forces of the information technology industry are focused on are designed to provide us with faster performance, greater speed and accelerated cadence for information processing. This pursuit of speed means we get more performant applications and the ability to execute quicker decision-making, with many of the information streams we now channel coming in real-time.

This velocity has consequences.

We need to think about the impact of faster web that drives ever-speedier clouds and the arrival of 5.5G (or advanced 5G as it is also known) mobile networks on the planet’s natural resources i.e. there’s no such thing as a free lunch so clearly we need to accommodate carbon neutral planning into development plans. But on a purely technical level, there’s also an increased data management responsibility i.e. faster clouds are characterized by many factors, but their speed means there is increased suitability for time-series databases in order for us to more accurately pinpoint what data happened when.

What is time-series data?

Just as they sound, time-series databases are built with time as a ‘key index’ factor in the creation of the algorithm that rules their behavior. Initially created to serve the needs of the financial trading business, as the fundamental conditions of computing have evolved with the popularization of cloud computing and the subsequent development of containerization (and Kubernetes container orchestration), the core cadence of our systems has continued to accelerate. Taking a time-series (also known as time-stamped) approach to data lifecycle management in the current age enables us to keep track of not just the who, what and where of data, but also the when – which hopefully gives us more insight into the why?

An increasingly vocal specialist in this space is KX, maker of kdb+ a time-series database and real-time analytics engine. The company’s recent work with its partner EnterpriseWeb (a telecom virtualization and automation specialist) to develop network observability and automation solutions for performance-critical systems gives us some insight into the way cloud computing is currently developing from a speed and self-optimizing perspective.

Bringing together complementary technologies, KX and EnterpriseWeb offer a closed-loop control mechanism (an automatic system engineered to maintain a desired state of operation without the need for human intervention) that can be deployed on any computing system or distributed network.

In terms of who brings what to the party, KX brings data services and real-time analytics capabilities built around its kdb+ time-series database for pattern recognition and the detection of anomalies. These feed EnterpriseWeb’s intelligent orchestration platform for intent-based network and service management which interprets the events, translates policies and takes actions to maintain a declared state for a self-scaling, self-healing, self-optimizing network. It’s what the companies call an integrated solution for a ‘zero-touch’ network and service management at scale.

An ‘engine’ for time-series

“We talk about KX as being an ‘engine’ for time-series and real-time analytics, powered by kdb+. As the results of [this] testbed project demonstrate, being the analytics engine in EnterpriseWeb’s intelligent orchestration platform can yield significant benefits. We’ve shown that using time series-based technology like KX can dramatically improve response times and increase network efficiency. We’re excited to see how our partnership can be applied to other use cases where the real-time monitoring and management of distributed networks and low latency response and remediation is a critical requirement.”

The two companies first teamed up as part of a team on a 5G edge network automation test-bed hosted by Intel Network Builders (INB), along with Red Hat, Keysight, Fortinet and Tech Mahindra. One of the distinguishing features of the testbed is the enhanced Self-Organizing Network (SON) to optimize the processing of secure [network data] packets, while significantly reducing energy consumption… so in fact, there is a planetary carbon-neutral consideration here after all.

An edge-optimized telco cloud

“Taking network services to the [computing] edge for low-latency customer use-cases requires extreme automation and efficiency so the explosion of new sites can be supported operationally,” said Dave Duggal, founder and CEO of EnterpriseWeb. “To streamline deployments, telcos will benefit from a common technology platform that can be flexibly configured for different industries, customers and use-cases so they can avoid being mired down in custom development projects. KX and EnterpriseWeb present a lightweight, low-latency, high-performance solution for an edge-optimized telco cloud platform.”

To date, self-organizing network implementations have generally been vendor-centric, narrow in scope, batch processed (i.e. not benefitting from real-time data streaming) and expensive to deploy and maintain. In contrast, this work has created technology that is rapid to deploy as an open, extensible, near-real-time SON that’s flexibly configurable for a wide variety of use cases.

Networks will continue to accelerate, so what matters now is precisely how we build the engineering mechanics required to enable an increasing amount of operations to happen autonomously, which is a large part of what’s happening here. The network backbone specialsist aren’t just listening to our needs, they’re keeping time too.

Waymo, Alphabet Inc.’s multibillion-dollar bet on self-driving cars and trucks, is pulling the human safety drivers out of its robotaxi test fleet in Los Angeles as it works to launch a commercial ride service in the second-biggest U.S. city.

The Mountain View, California-based company, which began testing electric Jaguar I-Pace crossovers loaded with cameras, lidar and other sensors in the sprawling city last October, will begin running them as “fully autonomous” cars in the next few weeks, initially in Santa Monica, Mauricio Peña, Waymo’s chief safety officer, told Forbes. The trials will occur initially in Santa Monica, with other neighborhoods to follow.

“We’re starting with a dense urban area and are going to expand and scale after we do this initial testing,” Peña told Forbes, without elaborating on when paid public rides might start. “We’re not providing a time line for Los Angeles right now, but are definitely doing all the steps we need to be ready to launch our commercial service.”

Progress for autonomous vehicles has been markedly slower than advocates imagined five or six years ago. A shakeout in the industry began last year when Ford and Volkswagen unexpectedly pulled the plug on Argo AI, their joint-venture autonomous tech startup, in October. Analysts and industry insiders expect more startups to fold this year as steep funding requirements and a longer-than-planned path to commercialization make it a less compelling bet. And though Waymo is among the most well-funded autonomous tech programs, it wasn’t spared from recent job cuts by Alphabet in recent weeks.

Still, with its paid ride service in the Phoenix area, the first of its kind in the U.S., and an upcoming commercial launch in San Francisco and L.A. to follow, Waymo is forging ahead. Still, aside from fine-tuning its AI-enabled software, the company can’t launch paid rides anywhere in California until the state’s Department of Motor Vehicles issues a permit allowing that.

In a test ride Forbes took with Waymo this month in Santa Monica, with a human backup driver behind the wheel, the vehicle showed improvement over earlier rides in suburban Phoenix, with no unexplained hard braking incidents or excessively cautious left turns. Still, the vehicle occasionally drew irritated honks from Angelenos driving behind, who wanted it to be less cautious at a couple of intersections. Those are the types of issues Waymo is working on daily, said Peña, who worked on spacecraft safety systems before joining Waymo.

“Safety is our priority; it’s our focus,” he said. “We also want to be good road citizens and be able to delight our riders, as well as keep them safe. That’s an area that we always pay attention to and are continuously improving.”

Meta is funding a national effort to help kids and teens get their nude or sexually explicit images removed from social media. Could the giant’s own move toward end-to-end encryption get in the way?

The National Center for Missing & Exploited Children has launched a platform, funded by Meta, to help kids and teens have naked or sexual photos and videos of themselves removed from social media.

The new service for minors, called Take It Down, was unveiled one year after the release of a similar tool for adults known as StopNCII (short for non-consensual intimate imagery). Today, Take It Down will work only across five platforms that agreed to participate: Meta-owned Facebook and Instagram, as well as Yubo, OnlyFans and Pornhub.

“We created this system because many children are facing these desperate situations,” said NCMEC’s president and CEO Michelle DeLaune. “Our hope is that children become aware of this service, and they feel a sense of relief that tools exist to help take the images down.”

How social media companies are protecting children and teens—and where they’re falling short—has emerged as a top bipartisan tech policy issue in the new Congress, and one where lawmakers stand the best chance at passing legislation. Still, there’s a long road ahead for any proposals to create or strengthen safeguards, and with no laws imminent, NCMEC continues to act as a bridge between online companies and law enforcement on child safety issues. In 2021, the nonprofit received nearly 30 million tips from across the industry about apparent sextortion and child sexual abuse material on their platforms. A majority were reported by platforms owned by Meta.

Minors and parents can use NCMEC’s free, anonymous tool to flag explicit material of an underage person that has been—or that they think will be—spread online. Once they’ve selected the sensitive photos or videos on their device, the platform generates the equivalent of a digital fingerprint of that photo (called a “hash”) that NCMEC then shares with the participating social media sites so they can try to find, restrict or remove, and keep monitoring for that content.

The photos or videos in question notably do not leave the device and are not viewed by anyone, according to NCMEC. (The hash also keeps victims unidentifiable both to NCMEC and to the social media app, signaling only that an image is harmful and should be taken down.) NCMEC said there have been no known breaches of its hash database that could potentially expose an individual’s identity.

But one of the biggest obstacles for this effort is the rise of end-to-end encryption. “When tech companies implement end-to-end encryption, with no preventive measures built in to detect known child sexual abuse material, the impact on child safety is devastating,” DeLaune, NCMEC’s president, told the Senate Judiciary Committee at a hearing this month. The elephant in the room is that the financial backer and most prominent company involved in Take It Down, Meta, is moving at a clip toward end-to-end encryption on both Instagram and Messenger.

How the two organizations can reconcile this is “the million dollar question,” NCMEC vice president Gavin Portnoy told Forbes, noting that NCMEC still does not know exactly what Meta’s plans are for encryption and what safeguards may be in place. (Meta said they’re a work in progress and that various encryption features are being tested on Instagram and Messenger.)

“We’re building Take It Down for the way the internet and the [electronic service provider] is situated today,” Portnoy said. “That really is a bridge we’re going to have to cross when we have a better understanding of what [Meta’s] end-to-end encryption environment looks like—and if there is a world where Take It Down can work.”

Meta’s global head of safety, Antigone Davis, told Forbes that “encryption will result in certain images not being identified by us” but that it’s becoming an industry standard in messaging that is vital to user privacy and security. Using hash technology and supporting NCMEC on projects like this one are among several ways Meta is trying “to really get in front of the problem,” she said. (Both parties declined to say how much money Meta has given to NCMEC to build Take It Down.)

“There’s no panacea for addressing these issues,” Davis said in an interview, describing Meta’s approach as “multi-faceted.” In addition to collaborating on this new system aimed at thwarting sextortion and the non-consensual spread of intimate images, Meta does not allow adult strangers to connect with minors through its messaging features, she said. It’s also using classifiers to prevent minors’ accounts and content from being pushed to adult Meta users who’ve exhibited suspicious behavior online, she added.

“I hope that people who’ve had images shared like that when they were young by anybody—by a parent who was exploiting them, by another adult who was exploiting them, by an ex-boyfriend or girlfriend—that they will feel interested to use this tool,” Davis said.

NCMEC said it has pitched some 200 other electronic service providers that have submitted information to its “CyberTipline” on joining the Take It Down program. It remains to be seen who else—beyond the small handful of social media players championing it today—will get on board. Yubo, for its part, said it hopes these early adopters will push others to join.

“It takes more than just government intervention to mitigate risk and solve issues online,” Yubo CEO Sacha Lazimi said over email. “All actors need to play their part. … Collaboration is key to identifying solutions to big challenges, so we do hope to see more platforms join the Take It Down initiative.”

Topline

The European Commission has banned staff from using TikTok on work devices, multiple media outlets reported Thursday, following similar prohibitions in the U.S. as Western governments head off mounting security concerns surrounding the Chinese-owned app amid souring relations with Beijing.

Key Facts

News Peg

TikTok has come under intense scrutiny from Western governments in recent months as officials grow increasingly concerned Beijing could tap the app, owned by Chinese company ByteDance, to spy on users. Though TikTok denies any accusation it would hand data over to authorities in Beijing, officials across the U.S. and Europe worry the Chinese government could force the company to comply. Multiple reports have shown ByteDance has already accessed U.S. user data on multiple occasions and planned to use the information to track American citizens, including Forbes journalists. The federal government and most state governments have already banned employees from using the app on official devices and U.S. lawmakers are seriously considering banning TikTok entirely to assuage concerns over national security. Other governments are wary and have urged caution when using the app. Politicians in some countries, such as the Netherlands, are considering bans of their own.

Big Number

32,000. That’s how many of the staff working for the European Commission are part of the European civil service, according to the EU. The decision could affect thousands in this group, which includes policy officers, researchers, lawyers and translators.

What To Watch For

The European Commission is one of three bodies overseeing lawmaking in the EU. The other two—the European Council and the European Parliament—are likely to consider bans of their own. The nature of the Parliament, made up of more than 700 elected officials from across the world’s largest trading bloc, means a ban could take much longer to implement.

Further Reading

TikTok Banned on EU Executive Staff Devices (WSJ)

How A TikTok Ban Would Work — And How TikTok Could Fight Back (Forbes)

Just over a year after initially announcing a partnership with lidar manufacturer Luminar, Mercedes-Benz has announced a major expansion of that relationship which includes multiple production programs. The new deal was announced as part of a broader Mercedes-Benz Operating System strategy update event streamed from the automaker’s North American R&D center in Sunnyvale, California.

Specific details remain scant for now, but more information is expected to be announced during a Luminar investor day at its Orlando, Florida, headquarters next week. Luminar will reveal more information about its upcoming product roadmap at that time. Currently, Luminar offers two products, its original Hydra lidar that’s intended just for development programs and the much lower-profile Iris, which is now in production.

The Mercedes-Benz program will utilize a next-generation version of the Iris sensor, which is expected to have a sleeker profile that should result in a less prominent hump at the top of the windshield. Luminar will almost certainly be focused on cost reduction with the next-generation sensor as it seeks to reach profitability while scaling volumes. The current Iris sensor is estimated to cost about $1,000 each. Innoviz, which won a production contract worth up to $4 billion from Volkswagen group in 2022 is targeting a price point of $500 or less for its next-generation InnovizTwo sensor, so Luminar is likely aiming for a similar price point.

The first series production application for the current Iris sensor is the SAIC Feifan R7 sedan that went into production in October 2022. Later in 2023, the same sensor will be standard equipment for the new Volvo EX90 and also available on the Polestar 3.

The current production programs are still relatively low in volume. However, the Volvo and Polestar could grow significantly if the brands opt to include standard lidar on upcoming electric models that share the same architecture. The Mercedes-Benz program is claimed to be a multi-billion dollar across a broad range of vehicle lines. Mercedes-Benz plans to utilize the Luminar sensors across a “broad range of next-generation production vehicle lines,” although there is no information about which vehicles at this time. Mercedes-Benz has had more than 2 million sales annually for the past several years, so this deal could easily amount to over 1 million units annually.

When Mercedes-Benz and Luminar announced their initial partnership in January 2022, that involved a production commitment for a single program. According to Luminar CEO Austin Russell, the new deal expands the potential volumes “by more than 10X.”

Mercedes-Benz plans to utilize the Luminar sensors as part of the assisted and automated driving system that it will deploy in the coming years based on the Nvidia Drive Hyperion platform. When the Mercedes-Benz Nvidia partnership was first announced in 2020, the first program was targeted at a 2024 launch, although given what has happened since then, some delay is probable. The new announcement only talks about a mid-decade launch which could be 2025 or later.

In order to support this expansion, Luminar will be growing its production capacity significantly. Luminar sources circuit boards and other major components from a number of contract manufacturing partners and does final assembly at its facility in Orlando, Florida, where it has capacity for about 250,000 units annually. To support the Mercedes-Benz program and other potential customers, it plans to add a new production facility somewhere in Asia along with an as-yet-unnamed partner.

While Mercedes-Benz is a premium vehicle manufacturer, this deal, along with the earlier Volkswagen-Innoviz deal, the GM-Cepton deal for Ultra Cruise and numerous other upcoming programs, indicate that lidar adoption will grow substantially over the remainder of this decade. That will likely drive costs down significantly and lead to further adoption on more mainstream models.