 |
|
|
|||||
In January, Meta announced the end of third-party fact-checkers on Facebook, Instagram, and Threads. The tech giant is betting on a new, community-driven system called Community Notes that draws on Xs feature of the same name and uses the Xs open algorithm as its basis. Meta is rolling out the feature on March 18. Anyone who wants to write and rate community notes can sign up now. The rollout will be throttled and, initially, notes wont appear publicly as Meta claims it needs time to feed the algorithm and ensure this system is working properly. The promise is enticing. A more scalable, less biased way to flag false or misleading content, driven by the wisdom of the crowd rather than the judgment of experts. But a closer look at the underlying assumptions and design choices raises questions about whether this new system can truly deliver on its promises. The concept, its UX implementation, and underlying technology surfaces challenges that, in my conversations with Meta’s designers, dont seem to have any clear, categorical answer. It feels more like a work-in-progress and not a clear-cut answer to the shortcomings of third-party fact-checking. Currently, Meta’s Community Notes are exclusively accessible on mobile devices within the Facebook, Instagram, and Threads apps. The mobile-first approach likely reflects the platform’s primary user base and usage patterns. Users who are eligible to contribute to Community Notes, after meeting specific criteria such as having a verified account and a history of platform engagement, can apply to be a contributor and add context to posts they believe contain misinformation (200,000 have already done so in the U.S., Meta tells me). Once in, they’ll find an option within the post’s menu to Add a Community Note. This triggers an overlay screen with a simple text editor that has a 500-character limit. The design also requires users to include a link, adding a layer of credibility to the note (although the link may not be a reliable source). [Images: Meta] Once a note is submitted, it’s evaluated by other Community Notes contributors. Meta uses an Xs open-source algorithmwhich may evolve later as they learn more about how it all really works, Meta saysto determine whether the note is helpful and unbiased. The algorithm considers various factors, like the contributor’s rating history and whether individuals who typically disagree with certain types of notes approve of it or not. Allegedly, the latter is the firewall to avoid coordinated activism against certain types of posts (although the algorithm hasnt been proved to be totally effective on X). The evaluation interface presents contributors with a clear and straightforward way to rate the note’s quality and helpfulness: a simple thumbs up/thumbs down system, which leads to another overlay menu in which they can select why they chose their option. Meta claims that if a note reaches a consensus among contributors with diverse viewpoints, it will then be publicly displayed beneath the original post, providing additional context without directly altering the post’s visibility or reach. The design aims to present the note as an informative supplement rather than a definitive judgment, allowing users to make their own informed decisions. [Images: Meta] An unsolvable problem? While the idea of crowdsourced fact-checking holds some theoretical appeal, Meta’s implementation appears to be riddled with the same vulnerabilities and unanswered questions that have affected X. On Elon Musk’s platform, Community Notes have failed to actually fact-check. They also suffer from extreme latency, or the amount of time that notes take to appear. A report from Bloomberg found that on average a typical note took seven hours to show up on the platform, but it can take as much as 70 hours, meaning false posts can go viral before they get checked. Community Notes on X have also failed to reduce engagement with false information. And because just 12.5% of Community Notes are seen, it denies their intrinsic value to the community. And lets not forget the potential to get gamed by particular interests. Metas own oversight board has pointed out huge problems with the plan. Still, the companys rationale to favor Community News over third-party fact-checkers hinges on two key arguments: scalability and reduced bias. Traditional fact-checking is a labor-intensive process that struggles to keep pace with the deluge of content on social media. That makes sense. By enlisting community members to flag and contextualize posts, Meta hopes to cover a much wider range of potentially problematic material. The social media company also argues that relying on a diverse group of contributors will mitigate the perceived bias of professional fact-checkers, who are often accused of political partisanship. The company and the designers cited a 2021 study by Allen et al. published in the scientific journal Science Advances titled Scaling up fact-checking using the wisdom of crowds as evidence that balanced crowds can achieve accuracy comparable to experts. Cracks in the foundation A critical examination of the study reveals a significant gap between the research and Meta’s proposed implementation. The study explicitly required political balancing of raters to achieve accurate results. Meta, on the other hand, has not clearly explained how it will ensure viewpoint diversity among contributors without collecting sensitive political data. In lieu of assessing a user’s past interactions on the platform, Meta plans to simply look at contributors’ past rating history in notes to assess whether or not a diversity of viewpoints has been achieved. Furthermore, the study only assessed the accuracy of headlines and ledes, not full articles. This raises serious concerns about the system’s ability to handle complex or nuanced misinformation, where the truth may lie in the details. The community notes limit of 500 characters adds to this concern. When I asked how it would be possible to truly add deep context to a post in which the truth is not binary (and lets face it: it almost never is), there wasnt a clear answer but a silence followed by the explanation that they could always expand the length if users demand it. Links to external sources can be included to provide more in-depth information, though they admitted that this adds another step for the reader to take. Its hard to imagine people clicking through in this era of content fast food. [Images: Meta] The company doesn’t have a plan for addressing one of the biggest issues with tackling misinformationwhich happens both via Community Notes and third-party fact-checking: the implied truth effect. Research shows that attaching notes to a subset of fake news headlines increases perceived accuracy of headlines without warnings. In the absence of these new notes, people might make the false assumption that a post is true. Meta’s designers say it will takes about the same time it takes on X for notes to go through the community fact-checking process, which means there will be plenty of time for fake news to go viral. Furthermore, X has shown that only a small percentage of posts get annotated, so the implied truth effect will, no doubt, be felt in Metas implementation of the same technologyat least in its current state. The old third-party fact-checking suffered from similar latency problems. No penalty Under the previous system, posts that fact-checkers identified as false or misleading had their distribution reduced. Community Notes, in contrast, will simply provide additional context, without impacting the reach of the original content. This decision flies in the face of research suggesting that warnings alone are less effective than warnings combined with reduced distribution. Meta says it wants to prioritize providing users with context rather than suppressing content. Its belief is that users can make their own informed decisions when presented with additional information. The fear is that demoting posts could lead to accusations of censorship and further erode trust in the platform. Meta says it will be monitoring the system, evaluating the latency, coverage, and the downstream effects of viewership and utilizing those metrics to guide future work, refinements, testing, and iterations. But Meta says there has been no A/B testing of Community Notes to see how it performs versus third-party fact-checking. Rather, the company is using this initial rollout phase as a public beta test, as a way to feed the algorithm with data from contributors so the system can get up and running. Fear, uncertainty, lots of doubt Twitter rolled out its proto version of Community Notes in 2020. Called Birdwatch, it continued to evolve with mixed results ever since Elon Musk took over and rebranded it with the current moniker. While Meta will use X’s open source algorithm as the basis of its rating system, feeding it with enough information to be operative could take quite a while. According to the Meta designers, the initial lack of public visibility is intended to allow them to train and thoroughly test the system and identify any potential problems before rolling it out to a wider audience. Meta isn’t saying how the notes will appear to all users, only pointing out in a press release that the plan is to roll out Community Notes across the United States once we are comfortable from the initial beta testing that the program is working in broadly the way we believe it should. Meta says it will gradually increase the visibility of the notes as it gains confidence in the system’s effectiveness, but did not provide a specific timeline or metrics for success. In a bid for transparency, Meta will release the algorithms that it uses. Its yet to be seen if Meta’s Community Notes will be more effective than the previous third-party fact-checking process. Nothing in the user experience suggests that it can solve the problems that X has had; logically, we can expect Meta to have many of the same issues, as well. In a historical moment where the truth is treated like malleable material, we could use a lot more certainties. Meta may have missed the chance to scientifically develop a new, non-derivative user experience that could avoid X’s problems. Instead, we are getting Musk’s broken toy with a coat of paint and the hope that, magically, it may work this time. Update 3/13/2025: An earlier version of this story stated Community Notes was launching on March 13. The feature is launching March 18.
Category:
E-Commerce
Identifying which companies you hope to work for is one of the biggest hurdles job seekers face. I know this because I was a search consultant for over 25 years. Now, I have an executive résumé and LinkedIn profile writing practice. And my clients almost always ask questions about how to find future employers. I advise them to use AI chatbot platforms like ChatGPT, Claude AI, and Perplexity. To help the platforms work their magic, I encourage them to use NAICS codes in their prompts. Heres how to do this: What are NAICS codes? NAICS is the acronym for the North American Industry Classification System. It assigns six-digit codes to companies as follows: The first two digits in a NAICS code identify economic sectors (e.g., 23 for Construction). NAICS has 20 sectors. The third and fourth digits divide economic sectors into subsectors. Example: 23 becomes 2382 for Building Equipment Contractors, a type of construction company. NAICS has 99 subsectors. The fifth and sixth digits divide economic subsectors into industries. Example: 2382 becomes 238220 for Plumbing, Heating, and Air-Conditioning Contractors, a type of building equipment contractor. NAICS has 1,000-plus industries. Thus, you can use NAICS codes at different levels to identify where you want to work. Once you know that, you can ask AI chatbot platforms to find companies in those NAICS codes. How AI chatbots can help find companies I asked ChatGPT how it finds companies. It searches for and analyzes public information from filings, directories, and the internet. It does in a minute or two what would take a job seeker hours, days, or weeks. I ran several searches on different platforms to show you how to use these chatbots to speed up your job search. You can see my prompts and results below. Prompts to find target companies I used these prompts to find companies by industry, location, and size: Prompt 1: Please list all the companies in NAICS code 713210 (Casinos) in Nevada. Claude AI provided a list of 55 large casinos. When I asked it to limit its results to Reno, it gave me 20 casino and gaming establishments. Prompt 2: List the 20 largest companies in the US in NAICS code 221115 (Wind Electric Power Generation). Perplexity listed 20 companies. When asked, it also shared locations, descriptions, and the 21st through 40th-largest companies. Prompt 3: List the companies in NAICS code 441110 (New Car Dealers) in Washington States King County. Perplexity named 17 dealerships, which was a good start but not comprehensive. ChatGPT wouldnt answer my query. Instead, it suggested I use Data Axle Reference Solutions, which I have recommended for years. DARS has a database of almost 100 million U.S. businesses. Its the ultimate resource if you hit a dead-end finding target employers, and its searchable by NAICS codes. Prompts to find recruiting, private equity, and venture capital firms Job seekers also want to find potential sources of opportunities, such as recruiting and private investment firms. To identify these targets, I used the following prompts. They included a subsector, industries, and specific investment strategies: Prompt 1: Please list search firms that recruit executives for companies in NAICS code 3254 (Pharmaceutical & Medicine Manufacturing). ChatGPT provided a list of 25 firms, although I had to re-prompt it with Any more? several times. Prompt 2: List venture capital firms that invest in AI start-ups (NAICS code 541745). ChatGPT provided a list of 28 firms. While I had to re-prompt it with Any more? several times, I stopped asking before it was done sharing firms. Prompt 3: Please list private equity firms that acquire turnaround clothing retailers (NAICS code 458110). ChatGPT provided a list of 17 firms. Again, I re-prompted it several times. Perplexity, prompted and re-prompted, gave me a list of 18 firms. You can use different platforms and variables at will. Doing so enables you to assemble lists of potential target companies in minutes.
Category:
E-Commerce
Across the U.S., dozens of proposed solar, wind, and battery projectsencompassing thousands of gigawatts of potential powerare backlogged as they wait to be allowed to plug into the power grid. And, even in areas where renewable energy projects are already online, their output is often heavily curtailed. This clean energy bottleneck stems from the fact that, as demand for renewable energy rises, the U.S. isn’t building new transmission lines fast enough to transport large amounts of clean energy from point A to point B. Now, theres a company looking to address that problem with a simple yet radical solution: Putting renewable energy into giant batteries and transporting those batteries by train. SunTrain is a San Francisco-based company founded by green energy developer Christopher Smith, who now serves as the companys president and chief technology officer. The idea, he explains, is to use the existing U.S. freight train systemwhich covers around 140,000 miles of terrainto bring renewable energy thats being curtailed by transmission bottlenecks to the areas that need it most. SunTrain is currently working on a pilot project that would run between Pueblo, Colorado, and Denver. If its approved by regulators, Smith says, he expects the pilot could be off the ground in just two years. [Photo: SunTrain] Current challenges to transporting renewables Clean energy is the fastest-growing source of electricity in the U.S. According to a report from American Clean power, 93% of the new energy capacity last year was solar, wind, and battery storage. The issue, Smith says, is that transmission line infrastructure lags far behind the rate of clean energy growth. The United States needs 300,000 miles of new transmission lines, like, right now. That’s the amount that we need immediately to keep up with current demand, Smith says. It’s also estimated that, to reach 100% renewables, as well as the electrification demand that we’ll have by 2050, well need over a million miles of new transmission lines by 2050. Currently, we’re building less than 1,000 miles a year. [Image: SunTrain] Building new transmission lines is challenging for a number of reasons, including environmental regulations, the time it takes, and the fact that any new lines would have to cross thousands of miles of privately owned land. In Colorado, for example, Smith says theres a lot of renewable energy in the states southeast corner, which flows through the grid to Pueblo. However, because theres not enough transmission line capacity between Pueblo and Denver, much of that power can’t ultimately be used. Once that renewable energy gets to Pueblo, there’s not enough transmission line capacity to get it into downtown Denver, Smith says. So that energy basically gets curtaileda fancy word for being wasted. Until now, the costly construction of new transmission lines has been the main solution that’s available. But Smith says this discussion overlooks a resource thats been used to transport energy for almost 200 years: railroads. The freight railroad network already moves virtually every single form of energy known to man that’s used in a real way: natural gas, coal, oil, ethanol, biomass, spent nuclear waste, various fossil fuels, and the list goes on and on, Smith says. So there’s this huge amount of overlap of our great railroad network and our electrical systems already. There is no reason why we cannot be moving battery trains over the freight rail network like we move every other form of energy. [Photo: SunTrain] SunTrain’s solution For its pilot project, SunTrain is partnering with Xcel Energy, Colorados largest electric utility. Xcel owns a coal plant in Pueblo (Comanche Generating Station) and a natural gas plant in Denver (Cherokee Generating Station) that are both set to be decommissioned within the next several years. Through a collaboration with SunTrain, these plants could potentially be re-powered with battery stored energy. Smith says SunTrain would use the existing substation inside Comanche Generating Stationwhich already has extensive railroad infrastructure from its history as a coa plantto charge its batteries from Pueblos bottlenecked grid. During the day, Smith says, the energy will likely be 100% renewable. Then, the batteries would be transported to Denver and the energy offloaded at the Cherokee Generating Station onto Denvers grid. (Charging and discharging take between four and six hours each, and the 139-mile trip from Pueblo to Denver takes about five hours by train.) A substation can turn energy into a format that can cover long distances without losing much energy, Smith says. A substation can also turn electricity generated from a power plant into a format that can be used by local homes and businesses. For SunTrain’s purposes, the Pueblo substation allows us to get the electricity formatted properly for our batteries while also collectively accessing all the various renewable energy generators in the region. [Photo: SunTrain] SunTrains proposed railcars will be made of 20-foot shipping containers, each of which will hold about 40 tons of batteries. The company designed proprietary charging and discharging systems that allow the energy to flow right from where it’s generated, whether it’s a solar array or a substation, right under the batteries on the railcar, Smith says. Then, once the train arrives at its destination, the discharging system would similarly allow the energy to flow right off the batteries. The whole process is designed so that the batteries never actually need to be removed from the train. [Image: SunTrain] In an interview with the podcast In the Noco, Smith said SunTrains first generation railcars are designed to match the freight railroads existing standards for coal trains, to ensure that the system itself doesnt need to change anything in order for SunTrain to come to market. Based on those parameters, each train will be built at between 8,000 and 9,000 feet long, with the capacity to carry around two gigawatt hours of power in total. Thats enough to power a city of 100,000 for a full day. Smith says the team has already tested a proof-of-concept train on several trips amounting to more than 10,000 miles on the Union Pacific network, traveling from SunTrains San Francisco testbed to discharge locations across California, Nevada, and Colorado. Now, the company is waiting for Colorados Public Utilities Commission to approve Xcels expenditure of about $125 million to begin construction on the pilot project. We tested the technology, the feasibility, made sure the mechanical standards were there, Smith says. Our manufacturing partners can deliver entire unit trains of thesemeaning 200 rail cars of batteries that could carry about 1.75 gigawatt hours of energy. So this isn’t something that’s far away, coming in the pipeline, or needing some kind of technological breakthrough. This is an immediately executable idea. It just needs the capital.
Category:
E-Commerce
All news |
||||||||||||||||||
|
||||||||||||||||||