70% of organizations struggle to provide completely connected user experiences across all channels. And almost three-quarters (72%) of organizations’ customer interactions are now digital.  MuleSoft 2022 Connectivity Benchmark Report MuleSoft’s 2022 Connectivity Benchmark Report found that 70% of organizations struggle to provide completely connected user experiences across all channels. At the same time, the report noted that almost three-quarters (72%) of organizations’ customer interactions are now digital.  Digitalization is accelerating, and organizations could lose on average nearly $7 million in revenue if they fail to complete digital transformation initiatives successfully. MuleSoft’s 2022 Connectivity Benchmark Report, in partnership with Vanson Bourne and Deloitte Digital, was produced from interviews with 1,050 IT leaders across the globe.  Here’s an executive summary of the report:  Organizations could lose on average $7m if they fail to successfully complete digital transformation initiatives. While companies have more applications than ever, they’re becoming less successful at integrating them. Legacy code and systems, siloed data, and skills shortages abound. Data silos are a persistent challenge for 90% of organizations. Integration demands are increasingly cited for more data scientists, business analysts, and customer support staff. The biggest challenges lie with incorporating dataderived insight into user-facing apps. Over half (52%*) of IT projects weren’t delivered on time last year (2021). The number of projects IT is being asked to deliver has increased by 40%* in the last year. Yet existing infrastructure continues to slow down project delivery speed, turning the department into an innovation bottleneck. Over half (55%) of organizations say it is difficult to integrate user experiences. This is a 7% increase from last year, driven by security and governance challenges, outdated infrastructure, and an inability to keep up with changing processes, tools, and systems. More than a quarter (26%) of business leaders now demand a company-wide API strategy. That’s almost double the figure from 12 months previous, although the number using APIs to build integrations has not changed in the intervening period. On average, over a third (35%) of organizations’ revenue is now generated by APIs and related implementations. In addition, they are recognizing how the combination of API-led connectivity and automation can deliver better employee and customer experiences.  Here are 13 powerful takeaways from the 2022 Connectivity Benchmark Report: Digital takes center stage: Nearly three-quarters (72%) of organizations’ customer interactions are now digital, and 93% say the speed at which projects take place is faster than it used to be 5 years ago. Digital takes center stage 2. The financial cost of slow digital transformation is significant: It’s calculated that they could lose on average nearly $7 million ($6,846,979) if they fail to successfully complete digital transformation initiatives. The business cost of slow digital transformation is significant  3. Adoption of apps is accelerating, but integration is falling behind: The average number an organization had this year was 976, a large increase from the figure a year ago (843). This could be an indication of growing shadow IT deployments in organizations during the pandemic. But more importantly, only 28% of these apps on average are integrated, down slightly from 29% in the 2021 report. The average lifetime for a typical application has also grown slightly to 4.1 years. Apps are on the rise, but integration is not 4. Firms are spending too much on custom integration: On average, companies spent $3.65 million in custom integration labor in 2021 versus $3.5 million in the previous 12 months. Firms are spending too much on custom integration: $9.5M in 2022 5. Data silos are persistent: Some 90% of respondents cite silos as a challenge, the same number as in the 2021 report, showing that little progress has been made in this crucial area. Data silos are causing significant business problems for most enterprises  6. Integration demands grow across the enterprise: Data scientists are most likely to have requirements for unlocking and integrating data (49%), followed by business analysts (44%), then customer support (42%). Integration demands grow across the enterprise The most pronounced integration challenges were: Incorporating data-derived insights into user-facing applications (75%) Reusing data sources across different user-facing applications (73%) Correlating data in the warehouse to derive insights (71%) Moving data from source systems into the data warehouse (70%) Also: Global research: 3 out of 4 professionals do not feel ready to work in a digital-first world 7. IT budgets continue to rise: 85% of responding organizations said this was true versus 77% in the 2021 report and 75% in the 2020 report. The number of projects IT is asked to deliver has increased by 40% in 2022, a jump from 30% in 2021. Over half (52%) of projects weren’t delivered on time last year. IT budgets are up, but so is demand 8. IT is struggling to deliver on time: On average, 52% of projects weren’t delivered on time over the past 12 months. However, the good news is that IT is more likely to be completing all the projects asked of them than in the period covered by the 2021 report: 44% versus 37%. IT is struggling to deliver on time  9. Organizations find it harder to integrate user experiences: Over half (55%) of organizations are finding it difficult to integrate user experiences. Nearly a third (30%) of organizations are able to provide completely connected user experiences across all channels. Connected experiences are more common today 10. Security and governance top integration challenges: Over half (55%) of organizations say it’s difficult to integrate user experiences, up from 48% a year ago. The biggest challenges are:  Security and governance (54%) Outdated IT infrastructure (46%) An inability to keep up with ever-changing processes, tools, and systems (42%) Data silos (42%) 11. Leadership now mandates a company-wide API strategy: The vast majority (98%) of organizations now use APIs. The number who mandate a company-wide API integration strategy has surged to 26% from 15% just a year ago. Nearly half (46%) of internal software assets are available for reuse, and 55% of organizations have a mature or very mature strategy for non-technical users.  APIs are virtually

With just weeks until Election Day, Vice President Kamala Harris and former President Donald Trump are in a dead heat among all registered voters: 48% say they would vote for Harris if the election were held today and 47% say they would support Trump. Another 5% of voters opt for a third-party candidate. A large majority of voters (82%) say they are certain they support Harris or Trump in the election. A much smaller share say they might change their mind about who to support – or only lean toward either of the two major party candidates when asked in a follow-up question. Among registered voters: 42% say they are certain about their support for Harris. 40% say they are certain about their support for Trump. 5% say they support Chase Oliver, Robert F. Kennedy Jr., Jill Stein or Cornel West. Roughly one-in-ten say they are likely to support or lean toward supporting Harris or Trump. When forced to choose between Harris or Trump in a two-way matchup, voters who initially say they would support a third-party candidate or do not lean toward any of the candidates are divided in their preferences: 35% say they would pick Trump if forced to choose, while 36% say Harris. About three-in-ten decline to answer. The snapshot of the race in this report is based on all registered voters. Not all voters will turn out in the 2024 presidential election – and those who are undecided or unsure about their choice are far less motivated to vote than are those who express more certainty. Voters who are certain about their choice are significantly more likely to say they have thought “a lot” about the candidates running for president, that it really matters who wins the race, and that they are extremely motivated to vote. Among Harris or Trump backers who say they’re certain about their candidate choice: 70% have thought “a lot” about the candidates. 84% say it really matters who wins. 70% say they are extremely motivated to vote. In contrast, far smaller shares of those who are unsure about supporting Harris or Trump say they are interested or invested in the outcome. Among voters who are less certain about their choice (i.e., those who support Trump or Harris, but say they could change their minds, or lean toward supporting Trump or Harris): 38% have given a lot of thought to the candidates. 36% say “it really matters” who wins. 19% are extremely motivated to vote. Those who support a third-party candidate are also significantly less likely to say they are motivated to vote, have thought a lot about the election, or that it “really matters” who wins. Voting preferences among registered voters The current snapshot of how all registered voters view the race shows Harris and Trump in a virtual tie: 48% say they would vote for or lean toward Harris if the election were held today, while 47% say they would support Trump. The underlying pattern in voting preferences mirrors those from last month: Harris is preferred among younger voters, Black voters and those with a college degree while Trump is favored among older voters, White voters and men. Gender Men are more likely to prefer Trump than Harris, while women favor Harris by a similar margin. Race and ethnicity More White voters prefer Trump to Harris (55% to 41%), while Black voters overwhelmingly back Harris (79% vs. 14%). Among all registered Hispanic voters, roughly half (54%) support Harris, while a smaller share (38%) support Trump. And Asian voters back Harris over Trump by two-to-one (62% vs. 30%). Education Registered voters without a four-year college degree back Trump over Harris (52% vs. 42%). The reverse is true for registered voters with a college degree (57% Harris vs. 38% Trump). Age Voters under 50 are more likely to prefer Harris over Trump (50% vs. 41%) while those ages 50 and older prefer Trump (52% vs. 46%). Voters under 50 are more likely to say they would support a candidate other than Harris or Trump (7%) than are those 50 and older (2%). For more on voting preferences among registered voters, refer to the detailed tables. 2020 turnout and support for Harris and Trump Trump has held onto most of his supporters from 2020 (94% say they support him this year). And Harris receives a similar level of support from President Joe Biden’s 2020 voters (92% support her). Just 5% of voters who backed Biden four years ago say they now support Trump, while 4% of voters who cast ballots for Trump last time now back Harris. Voters who did not vote in 2020 – either because they chose not to or were too young – are split about evenly between Trump (45%) and Harris (44%), while 8% say they back a third-party candidate. 2020 Trump voters Trump’s support is strongest among voters who backed him in both 2016 and 2020 (97%) and among those who did not vote in 2016 but turned out for him four years ago (93%). Most voters who switched from Clinton or another candidate in 2016 to Trump in 2020 continue to support Trump (78% back him now), but their support is lower than among his other 2020 voters. About one-in-ten of these voters (11%) say they support a third-party candidate, while an identical share supports Harris. 2020 Biden voters Support for Harris among Biden’s 2020 voters follows a similar pattern: 95% of voters who voted for Clinton in 2016 and Biden in 2020 say they support Harris, and 87% of those who did not vote in 2016 but cast a ballot for Biden in 2020 now support Harris. About three-quarters of Biden’s 2020 voters who backed Trump or another candidate in 2016 are with Harris this year (76%), while 16% say they support Trump and 4% support a third-party candidate. Harris and Trump supporters’ views of their vote Harris’ supporters are more likely than Biden’s supporters were four years ago to view their vote as more

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More The number of AI video generation models continues to grow with a new one, Pyramid Flow, launching this week and offering high quality video clips up to 10 seconds in length — quickly, and all open source. Developed by a collaboration of researchers from Peking University, Beijing University of Posts and Telecommunications, and Kuaishou Technology — the latter the creator of the well-reviewed proprietary Kling AI video generator — Pyramid Flow leverages a new technique wherein a single AI model generates video in stages, most of them low resolution, saving only a full-res version for the end of its generation process. It’s available as raw code for download on Hugging Face and Github, and can be run in an inference shell here but requires the user to download and run the model code on their own machine. NEW: Open Source Text/ Image to video model is out – MIT licensed – Rivals Gen-3, Pika & Kling 🔥 > Pyramid Flow: Training-efficient Autoregressive Video Generation method> Utilizes Flow Matching> Trains on open-source datasets> Generates high-quality 10-second videos>… pic.twitter.com/ZU0X6YMxWr — Vaibhav (VB) Srivastav (@reach_vb) October 10, 2024 At inference, the model can generate a 5-second, 384p video in just 56 seconds—on par with or faster than many full-sequence diffusion counterparts — though Runway’s Gen 3-Alpha Turbo still takes cake in terms of speed of AI video generation, coming in at under one minute and often times 10-20 seconds in our tests. We haven’t had a chance to test Pyramid Flow yet, but the videos posted by the model creators appear to be incredibly lifelike, high enough resolution, and compelling — analogous to those of proprietary offerings. You can see various examples here on its Github project page. Indeed, Pyramid Flow is available designed now to download and use — even for commercial/enterprise purposes — and is designed to compete directly with paid proprietary offerings such as Runway’s Gen-3 Alpha, Luma’s Dream Machine, Kling, and Haulio, which can cost hundreds of even thousands of dollars a year for users on unlimited generation subscriptions. As the race between various AI video providers to gain users continues, Pyramid Flow aims to bring more efficiency and flexibility to developers, artists, and creators seeking advanced video generation capabilities. A new technique for high-quality AI videos: ‘pyramidal flow matching’ AI video generation is a computationally intensive task that typically involves modeling large spatiotemporal spaces. Traditional methods often require separate models for different stages of the process, which limits flexibility and increases the complexity of training. Pyramid Flow is built on the concept of pyramidal flow matching, a method that drastically cuts down the computational cost of video generation while maintaining high visual quality, completing the video generation process as a series of “pyramid” stages, with only the final stage operating at full resolution. It’s described in a pre-reviewed paper, “Pyramidal Flow Matching for Efficient Video Generative Modeling,” submitted to open access science journal arXiv on October 8, 2024. The authors include Yang Jin, Zhicheng Sun, Ningyuan Li, Kun Xu, Hao Jiang, Nan Zhuang, Quzhe Huang, Yang Song, Yadong Mu, and Zhouchen Lin. Most of these researchers are affiliated with Peking University, while others are from Kuaishou Technology. As they write, the ability to compress and optimize video generation at different stages leads to faster convergence during training, allowing Pyramid Flow to generate more samples per training batch. For example, the proposed pyramidal flow reduces the token count by a factor of four compared to traditional diffusion models, which results in more efficient training. The model can produce 5- to 10-second videos at 768p resolution and 24 frames per second, all while being trained on open-source datasets. Specifically, the paper states that Pyramid Flow was trained on trained on: LAION-5B, a large dataset for multimodal AI research. CC-12M, a dataset of web-crawled image-text pairs. SA-1B, which features high-quality, non-blurred images. WebVid-10M and OpenVid-1M, which are video datasets widely used for text-to-video generation. In total, the authors curated approximately 10 million single-shot videos. However, many of these “public” or “open source” datasets have in recent years come under fire from critics for including copyrighted material without permission or informed consent of the copyright holders, and LAION-5B in particular accused of hosting child sexual abuse material. Separately, Runway is among the companies being sued by artists in a class action lawsuit for training on materials without permission, compensation, or consent — allegedly in violation of U.S. copyright. The case remains being argued in court, for now. Permissively licensed, open source for commercial usage Pyramid Flow is released under the MIT License, allowing for a wide range of uses, including commercial applications, modifications, and redistribution, provided the copyright notice is preserved. This makes Pyramid Flow an attractive option for developers and companies looking to integrate the model into proprietary systems, and could challenge Luma AI and Runway as both look to offer paid application programming interfaces for developers seeking to integrate their proprietary AI video generation technology into customer or employee-facing apps. Yet those proprietary models already exist as inferences suitable for developers, while Pyramid Flow has a demo inference on Hugging Face, it is not suitable for building full applications atop it and users would need to host their own version of an inference, which could also be costly, despite the model itself being “free.” In addition, Pyramid Flow may prove to be enticing to film studios looking to leverage AI to gain efficiencies, cut costs, and explore new creative tools. One major film studio, Lionsgate — owner of the John Wick and Twilight films franchises, among many other tiles — recently inked a deal for an unspecified sum with Runway to train a custom AI video generation model. Furthermore, Titanic and Terminator director James Cameron joined the board of AI video and image model provider Stability (the latter also subject to the same class-action lawsuit from artists as Runway). Using Pyramid Flow,

Image: Sergei Supinsky/AFP/Getty Images The list of companies choosing to suspend operations in Russia continues to grow amid Russia’s invasion of Ukraine, with Netflix among the latest to join that list. Last week, the streaming service said it was pausing all future projects and acquisitions in Russia. But now the company is taking an additional step and is shutting down its service in the country entirely. “Given the circumstances on the ground, we have decided to suspend our service in Russia,” a Netflix spokesperson told ZDNet. Meanwhile, TikTok has announced that it will be suspending any livestreaming and new content on its video service “in light of Russia’s new ‘fake news’ law”. According to the Moscow Times, Russian President Vladimir Putin signed a new law on Friday that bans what the country calls “fake” news about the military. This law will target any statements referring to the invasion of Ukraine as an “invasion”, any attempt to discredit the armed forces, or calls for sanctions on Russia. Those found to be spreading so-called fake news could face up to 15 years in prison. “TikTok is an outlet for creativity and entertainment that can provide a source of relief and human connection during a time of war when people are facing immense tragedy and isolation. However, the safety of our employees and our users remain our highest priority,” the company announced in a tweet. TikTok’s in-app messaging service will not be affected, the company assured. “We will continue to evaluate the evolving circumstances in Russia to determine when we might fully resume our services with safety as our top priority,” TikTok added. Mastercard and Visa also announced on Saturday they were suspending operations in Russia. Both had already blocked multiple financial institutions from using their respective payment networks as directed by regulators globally. Under these latest steps, cards issued by Russian banks are no longer supported by the Mastercard or Visa networks. Plus, any Mastercard or Visa issued outside of the country will not work at Russian merchants or ATMs. “We are compelled to act following Russia’s unprovoked invasion of Ukraine, and the unacceptable events that we have witnessed,” Visa chairman and CEO Al Kelly said in a statement. “We regret the impact this will have on our valued colleagues, and on the clients, partners, merchants and cardholders we serve in Russia. This war and the ongoing threat to peace and stability demand we respond in line with our values.” Further, DXC said it will be exiting Russia completely. “Based on the aggression from the Russian government, we are no longer pursuing business in Russia and have committed to exit this market,” the company stated. The company currently has approximately 4,000 staff in Russia and said it will be supporting them during this period. “Currently, our number one priority is to care for our colleagues in the region. Our dedicated DXC team is working around the clock to provide shelter, financial assistance, health care, and relocation support to our colleagues and their families,” the company said. DXC said it will also double any employee donations to the Red Cross humanitarian efforts, which includes providing direct financial support to impacted colleagues. “We will continue to assess the situation and take the required measures leveraging the strength of our global team to help minimise any impact on our DXC colleagues and customers,” the company said. source

Samsung on Monday confirmed that the company recently suffered a cyberattack, but said that it doesn’t anticipate any impact on its business or customers. Last week, South American hacking group Lapsus$ claimed it had stolen 190GB of confidential data, including source code, from the South Korean tech giant’s servers. The group also posted snapshots of the alleged data online. Samsung has now confirmed in a statement, without naming the hacking group, that there was a security breach, but it asserted that no personal information of customers was compromised. SEE: DDoS attacks that come combined with extortion demands are on the rise “We were recently made aware that there was a security breach relating to certain internal company data. Immediately after discovering the incident, we strengthened our security system,” the company said. “According to our initial analysis, the breach involves some source codes relating to the operation of Galaxy devices, but does not include the personal information of our consumers or employees. Currently, we do not anticipate any impact to our business or customers. We have implemented measures to prevent further such incidents and will continue to serve our customers without disruption.” On whether the company had received a demand for payments or was in negotiation to do so with any hacking group, a company spokesperson declined to comment on the matter. Hacking group Lapsus$ also claimed last month that it had stolen 1TB data from GPU giant Nvidia, while also posting snapshots of some of the data online. In response last week, Nvidia also confirmed that some employee credential and proprietary information was stolen, but also said it doesn’t expect disruption to its business. source

There is no single platform or pathway that most Americans use as their primary way to get news about the election. About a third of U.S. adults say television is the most common way they get political and election news (35%), while roughly one-in-five each most commonly get it from news websites or apps (21%) or social media (20%). Smaller shares turn primarily to search engines like Google (8%) or to radio or podcasts (5% each). And just 3% of U.S. adults say print newspapers or magazines are their most common way of getting election news. The platforms Americans use for election news differ by age Different age groups turn to these platforms at drastically different rates. Almost half of U.S. adults under 30 (46%) most commonly use social media for political news. This is twice the share of 30- to 49-year-olds who say the same (23%). And it far surpasses the small shares of those ages 50 to 64 (7%) and 65 and older (3%) who primarily use social media in this way. Older adults are far more likely than their younger counterparts to say TV is their most common way of getting political news. TV is the most common source of political news for the majority of Americans ages 65 or older (63%). By comparison, just one-in-ten adults under 30 say TV is their primary source for this type of news. Americans ages 30 to 49 have the most varied habits when it comes to election news. People in this group are about equally likely to say they get most of their election news from TV (23%), social media (23%) and news websites or apps (24%). Republicans and Democrats are mostly similar in which platforms they use for political news. For example, about a third of Republicans and Republican leaners get political news from television (34%), identical to the share of Democrats and Democratic leaners who say the same. Which news organizations Americans turn to for election news The survey also asked respondents whether they use eight specific news outlets (or groups of outlets) as major or minor sources of political and election news. For most of these outlets, the share of Americans who use each for political news has remained largely steady since we last asked these questions ahead of the 2020 election.  Network television news remains the most widely used of these sources: 68% of Americans say ABC, CBS or NBC is at least a minor source for election news, including three-in-ten who name this as a major source. About half of U.S. adults use CNN (53%) or Fox News (50%) as a major or minor source of political news, while about four-in-ten say the same about MSNBC (41%). Both The New York Times (38%) and The Washington Post (32%) remain sources of election news for many Americans, although the shares who turn to these traditional newspapers have declined modestly since 2020 (from 42% and 39%, respectively). About a third of Americans name NPR (34%) as at least a minor source of election news, while 27% use conservative talk radio, such as Sean Hannity or Mark Levin (27%). The share who cite talk radio as at least a minor source of election news is identical to 2020, when the question named Rush Limbaugh as an example before his death in 2021. How election news sources differ by political party Republicans and Democrats differ substantially in the rates that they use all eight major news brands our survey asked about to get political and election news. Two of these sources – Fox News and conservative talk radio – are much more commonly used by Republicans. For example, while 69% of Republicans say Fox News is at least a minor source of political and election news for them, only 32% of Democrats say the same. But Democrats are more likely than Republicans to use each of the other six sources, ranging from cable networks like CNN and MSNBC to network TV channels, legacy newspapers and NPR. For instance, 48% of Democrats say they use NPR as a major or minor source of election news, compared with 21% of Republicans. Age differences within the parties Younger Republicans are less likely than their older GOP counterparts to say they use both Fox News and talk radio as sources for political and election news. For example, about six-in-ten Republicans under 30 (58%) say Fox News is at least a minor source of election news for them, compared with almost eight-in-ten Republicans ages 65 or older (79%) who say the same. By contrast, younger Republicans are more likely than older Republicans to use several other sources, including The New York Times, The Washington Post and CNN. Younger Democrats also are more likely than older Democrats to use The New York Times and The Washington Post as major or minor sources of political news during this election year. For instance, two-thirds of Democrats under 30 say they use The New York Times, compared with just 38% of Democrats ages 65 and older. Compared with their elders, younger Democrats are less likely to use the network TV stations or MSNBC as sources of election news – in line with the broader pattern that older Americans are more likely to name TV as their most common source for such news. But Democrats of different age groups use CNN as a source of election news at nearly identical rates. source

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More Fresh off the news of its record-setting $6.6 billion funding round, OpenAI is updating its signature AI app ChatGPT with a new feature — Canvas — that allows users to see, directly edit, and easily modify just selected portions of the chatbot’s outputs in a side-by-side panel view. The feature, built atop OpenAI’s GPT-4o model, can also give suggestions and implement changes to the responses in the right-side panel without needing to output an entirely new response. It seems like a direct challenge to rival AI startup Anthropic’s Artifacts feature for its chatbot Claude, which launched in June 2024. It also offers a side panel view to display and run simple Python programs that update based on the user’s prompts. Canvas is rolling out to ChatGPT Plus and Teams users, with subscribers to Enterprise and Edu tiers following in a few days. Following the beta, OpenAI plans to make the feature available to all ChatGPT users.  Users on the social network X predicted Canvas before its initial release and speculated that OpenAI would soon release the feature, which turned out to be correct. What ChatGPT Canvas is good for Daniel Levine, product manager for Canvas, told VentureBeat in an interview that sometimes, the vertical top-to-bottom chat window is too limited for some of the most common use cases. “We know a lot of people use ChatGPT for writing and coding, those are two of the top use cases we see,” Levine said. “But the chat interface is a bit limiting, especially for projects where you want revisions or editing. There’s a lot of back and forth, and comparing changes is hard, so that’s where Canvas steps in.” Levine is talking about prompting ChatGPT to edit its responses. Without Canvas, when users ask ChatGPT to generate a draft email and feel the response needs to be longer, funnier, or friendlier, they have to prompt ChatGPT again, and rewrites appear in the same conversation. Sometimes, the draft changes drastically; sometimes, the change is subtle, but often, you have to scroll back and forth to double-check what changed.  OpenAI hopes that with Canvas, the process will be easier. Users can, of course, reprompt ChatGPT, but if they only want a few words changed, they don’t have to copy text that is still very much a first draft to a different document and edit it themselves.  How to use Canvas Canvas will open after users toggle the model picker, where they normally choose which version of OpenAI’s models to use. ChatGPT can also detect when Canvas would be appropriate to open or when the user sends the prompt,” Use Canvas.” They can then prompt ChatGPT to either write something or generate code. For example, if a user wants to write an email to a colleague, they prompt ChatGPT with the request, and it opens the Canvas window with the text it just generated.  Users can continue asking ChatGPT to refine the text, which will reflect on the Canvas screen. People can also edit directly on Canvas, or give instructions to the chatbot by highlighting some text. On each Canvas window, there is a set of shortcuts for users to click that can adjust text length, the reading level, add emojis or even give a final polish. ChatGPT can also provide suggestions for the text, which will show up on Canvas in text boxes, similar to how comments look on Google Docs. It can also translate into supported languages.  Canvas looks different based on the task. The writing Canvas looks like a Word document, while the coding Canvas includes line numbers for easier code editing. Shortcuts for coding will also be different. Users can review code, fix bugs, add logs and comments and port to a different coding language through Canvas.  Directly competing with Anthropic Claude’s Artifacts feature Canvas’s separate window and model collaboration call to mind another chatbot with a window that allows people to clearly see any changes made through new prompts: Anthropic’s Claude Artifacts. VentureBeat’s Michael Nunez reported that Claude Artifacts makes accessible and easy-to-understand interfaces an essential feature in chatbots and called it “this year’s most important AI feature.” Unlike Canvas, Artifacts is already generally available to all Claude chat users.  Artifacts also let people see how their generated code looks, as they let users prompt Claude to write code, edit it and then see the fruits of the labor, like a prototype website or game. Canvas just shows users the generated code and the edits around it.  The new interface battleground headed by OpenAI’s Canvas and Anthropic’s Artifacts does point to a problem smaller, third-party AI applications have been trying to solve, which is how to make chatbots easier to read and use. Apps such as Hyperwrite, Jasper and JotBot all generate and offer text editing. Many other software include similar editing features for both code and text with the idea that users don’t need to leave the chat window to make changes to their work.  Eventually, features like Canvas and even Artifacts may become commonplace as more people want to work with chatbots more streamlined and collaboratively.  “We do think collaborative work is an important part of the workplace,” Levine said. “So we’re taking a first step here.” source

Brazilian startup Beegol has raised $4.2m in a Series A round for its machine learning powered diagnostic and self-healing technology aimed at cost savings for network service providers. Beegol, onomatopoeically named after the Beagle dog, can sniff out ISP network problems and in many cases, it can make repairs with a remote technology embedded in the firmware chips of most modems.  The company’s goal is to improve reliability of Internet based services and allow Internet Service Providers to improve reliability while saving money on rolling out repair trucks. “It’s important for operators to have machine learning and diagnostics so that they understand root causes of these outages so they can prevent it from happening again or determine if it’s their issue or it could be a power line issue,” says Charles Hong, Chief Strategy Officer and co-founder of Beegol.  “It could be an issue affecting, for example, a Zoom server if our communications went out just now.” The increased complexity of networks raise the number of points of potential failure and they raise the complexity in diagnosing problems. Without good diagnostics ISPs and telcos cannot improve reliability. The company’s market advantage is that it managed to get its software into the firmware of the dominant chipsets used in manufacturing modems. The firmware software provides unparalleled advantages in terms of diagnostic tools and remote repairs. The data collection allows operators to take advantage of the technology’s machine learning features to better understand problems and where future issues are likely to arise.  The machine learning algorithms develop specific use cases for each problem. Gilberto Mayor, CEO and co-founder, explains, “When you are watching Netflix on iPad, your iPad should work well, your Wi-Fi should work well. There should be no interference from your neighbor’s channel. You should be close to the Wi-Fi access point. The local access should work well. The transport network should work well.” The user is likely to call their Internet provider regardless of who is responsible and dissatisfaction with service is a top reason for switching providers. Beegol allows ISPs to quickly find out about a problem and start fixing it before any customer calls. ISPs can also remotely initiate some repair functions from within the firmware. Beegol recently announced that its remote diagnostic technology will be integrated into the open industry standard Reference Design Kit (RDK), which is ubiquitously used by nearly all video and Internet network providers to manage network services. It has also joined RDK’s Technical Advisory Board. Beegol already has customers among large ISPs and unlike many Series A startups, it already has revenues of more than $8 million and about 43 employees. Hong said that the company waited several years to raise money to retain greater ownership. Sometimes, a self-sustained startup will take a venture capital investment, not because they need the money, but because they hope to get some publicity from publications that cover their industry. This helps provide some validity among potential customers and future investors. The Series A funding was provided by Indicator Capital, a Brazilian VC firm that manages a $45m fund raised last year mostly from Qualcomm and the Brazilian Development Bank. Indicator said that it was attracted to Beegol because of the importance of trends such as working from home, video gaming, and new online social environments.  The Beegol investment is one of many expected this year in the Latin American region, which continues to lead as the top hot spot for global VC funding. Brazil dominates the Unicorn list of startups valued at over $1 billion but other countries such as Colombia and Chile are catching up. source

Most Americans are following the election closely, though roughly half say it’s hard to know what’s true People watch the ABC News presidential debate between Vice President Kamala Harris and former President Donald Trump at a watch party in West Hollywood, California, on Sept. 10, 2024. (Mario Tama/Getty Images) Pew Research Center conducted this study to better understand how U.S. adults get news and information about the 2024 presidential election. The survey of 9,680 U.S. adults was conducted from Sept. 16 to 22, 2024. Everyone who completed the survey is a member of the Center’s American Trends Panel (ATP), a group of people recruited through national, random sampling of residential addresses who have agreed to take surveys regularly. This kind of recruitment gives nearly all U.S. adults a chance of selection. Surveys were conducted either online or by telephone with a live interviewer. The survey is weighted to be representative of the U.S. adult population by gender, race, ethnicity, partisan affiliation, education and other factors. Read more about the ATP’s methodology. Here are the questions used for this report, the topline and the survey methodology. Pew Research Center is a subsidiary of The Pew Charitable Trusts, its primary funder. This is the latest report in Pew Research Center’s ongoing investigation of the state of news, information and journalism in the digital age, a research program funded by The Pew Charitable Trusts, with generous support from the John S. and James L. Knight Foundation. As Americans navigate the abundance of information about the 2024 U.S. presidential election, many of them are seeing news they think is inaccurate – and having a hard time distinguishing fact from fiction. A new Pew Research Center survey finds that about three-quarters of U.S. adults (73%) say they have seen inaccurate news coverage about the election at least somewhat often, including 37% who have seen this kind of information extremely or very often. About half of Americans (52%) say they generally find it difficult to determine what is true and what is not when getting news about the election. And 28% separately say that it’s been difficult for them to find reliable information about the presidential election. Supporters of the two major political parties are having vastly different experiences with election news in these ways. Republicans (including independents who lean toward the GOP) are much more likely than Democrats and Democratic leaners to say they have seen inaccurate coverage and that they are having a hard time sorting out the truth. And Republicans are less likely than Democrats to say it is easy for them to find reliable information about the election (29% vs. 52%, respectively). Overall, Americans are much less likely to be wary of the information from their most frequent sources of election news. Much smaller shares say they at least sometimes see inaccurate information from the sources they turn to most often for news and information about the election than say the same about election news in general. At the same time, people also say inaccurate information is coming from sources other than news outlets: 58% of Americans say they have heard people they know share information about the election that they think is inaccurate at least somewhat often. There is little difference between the parties on this question – Republicans and Democrats say they hear inaccurate information from their personal circles at roughly the same rates. Jump to more information about Americans’ views on the accuracy of election news. How Americans’ attention toward election news has changed throughout the campaign Overall, about seven-in-ten Americans say they are following news about the candidates for the 2024 presidential election very (28%) or fairly (40%) closely. This is up from 58% in April and 65% in July, mirroring a pattern also seen in 2020 in which attention increases closer to Election Day. Between April and September of this year, the increase was driven particularly by Democrats, who were slightly less likely than Republicans to be following election news closely in the two earlier surveys. Now, there is no difference between the two parties in their levels of attention. About six-in-ten Americans say they are worn out by so much coverage of the campaign, while nearly four-in-ten say they like seeing a lot of coverage. These shares are similar to those measured in April, as well as in surveys conducted in 2020 and 2016. Jump to more about how much attention Americans are paying to election news. What kinds of news about the election are Americans seeing – and what do they want to see? The survey asked Americans how much they have heard about several specific election-related topics that have been in the news in recent months. We found that most people say they have heard or read a lot about Vice President Kamala Harris replacing President Joe Biden as the Democratic nominee (70%), the first assassination attempt on former President Donald Trump in July (66%), and the debate between Harris and Trump on Sept. 10 (64%). (The survey questions were finalized before the second assassination attempt on Trump in September.) Smaller shares say they have heard a lot about the vice presidential candidates, the two parties’ national conventions, and third-party candidate Robert F. Kennedy Jr. endorsing Trump when he dropped out of the race. Still, large majorities of Americans say they have heard at least a little about each of these topics. More generally, news about the candidates’ comments and actions on the campaign trail appears to be the most common type of coverage people are seeing. Four-in-ten U.S. adults say this is the type of election news they see most, outpacing information about the candidates’ stances on issues (17%), their moral characters (14%) or which candidate is leading the race (13%).   This does not align with what Americans are most interested in. For example, higher shares say they are extremely or very interested in news about candidates’ stances on the issues (75%) and their moral characters (60%) than say the same about

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More A trio of scientists consisting of Demis Hassabis, co-founder and CEO of Google’s AI division DeepMind, as well as John Jumper, Senior Research Scientist at Google DeepMind and David Baker of the University of Washington have been awarded the 2024 Nobel Prize in Chemistry for their groundbreaking work in predicting and developing new proteins. The DeepMinders won for AlphaFold 2, an AI system launched in 2020 capable of predicting the 3D structure of proteins from their amino acid sequences. Meanwhile, Baker won for leading a laboratory where the 20 amino acids that form proteins were used to design new ones, including proteins for “pharmaceuticals, vaccines, nanomaterials and tiny sensors,” according to the Nobel committee’s announcement. The award highlights how artificial intelligence is revolutionizing biological science — and comes just one day after what I believe to be the first Nobel Prize awarded to an AI technology, that one for Physics to fellow Google DeepMinder Geoffrey Hinton and Princeton professor John J. Hopfield, for their work in artificial neural networks. The Royal Swedish Academy of Sciences announced the prize as it did with the Physics one, valued at 11 million Swedish kronor (around $1 million USD), split among the laureates — half will go to Baker and the other half divided again in fourths of the total to Hassabis and Jumper. Breakthrough on a biology problem unsolved for half of a century The committee emphasized the unprecedented impact of AlphaFold, describing it as a breakthrough that solved a 50-year-old problem in biology: protein structure prediction, or how to predict the three-dimensional structure of a protein from its amino acid sequence. For decades, scientists knew that a protein’s function is determined by its 3D shape, but predicting how the string of amino acids folds into that shape was incredibly complex. Researchers had attempted to solve this since the 1970s, but due to the vast number of possible folding configurations (known as Levinthal’s paradox), accurate predictions remained elusive. AlphaFold, developed by Google DeepMind, made a breakthrough by using AI to predict the 3D structures of proteins with near-experimental accuracy, meaning that the predictions made by AlphaFold for a protein’s 3D structure are so close to the results obtained from traditional experimental methods—like X-ray crystallography, cryo-electron microscopy, or nuclear magnetic resonance (NMR) spectroscopy—that they are almost indistinguishable. When AlphaFold achieved “near-experimental accuracy,” it was able to predict protein structures with a level of precision that rivaled these methods, typically within an error margin of around 1 Ångström (0.1 nanometers) for most proteins. This means the model’s predictions closely matched the actual structures determined by experimental means, making it a transformative tool for biologists. Hassabis and Jumper’s work, developed at DeepMind’s London laboratory, has transformed the fields of structural biology and drug discovery, offering a powerful tool to scientists worldwide. “AlphaFold has already been used by more than two million researchers to advance critical work, from enzyme design to drug discovery,” Hassabis said in a statement. “I hope we’ll look back on AlphaFold as the first proof point of AI’s incredible potential to accelerate scientific discovery.” AlphaFold’s Global Impact AlphaFold’s predictions are freely accessible via the AlphaFold Protein Structure Database, making it one of the most significant open-access scientific tools available. Over two million researchers from 190 countries have used the tool, democratizing access to cutting-edge AI and enabling breakthroughs in fields as varied as molecular biology, drug development, and even climate science. By predicting the 3D structure of proteins in minutes—tasks that previously took years—AlphaFold is accelerating scientific progress. The system has been used to tackle antibiotic resistance, design enzymes that degrade plastic, and aid in vaccine development, marking its utility in both healthcare and sustainability. John Jumper, co-lead of AlphaFold’s development, reflected on its significance, stating, “We are honored to be recognized for delivering on the long promise of computational biology to help us understand the protein world and to inform the incredible work of experimental biologists.” He emphasized that AlphaFold is a tool for discovery, helping scientists understand diseases and develop new therapeutics at an unprecedented pace. The Origins of AlphaFold The roots of AlphaFold can be traced back to DeepMind’s broader exploration of AI. Hassabis, a chess prodigy, began his career in 1994 at the age of 17, co-developing the hit video game Theme Park, which was released on June 15 that year. After studying computer science at Cambridge University and completing a PhD in cognitive neuroscience, he co-founded DeepMind in 2010, using his understanding of chess to raise funding from famed contrarian venture capitalist Peter Thiel. The company, which specializes in artificial intelligence, was acquired by Google in 2014 for around $500 million USD. As CEO of Google DeepMind, Hassabis has led breakthroughs in AI, including creating systems that excel at games like Go and chess. By 2016, DeepMind had achieved global recognition for developing AI systems that could master the ancient game of Go, beating world champions. It was this expertise in AI that DeepMind began applying to science, aiming to solve more meaningful challenges, including protein folding. The AlphaFold project formally launched in 2018, entering the Critical Assessment of protein Structure Prediction (CASP) competition—a biannual global challenge to predict protein structures. That year, AlphaFold won the competition, outperforming other teams and heralding a new era in structural biology. But the real breakthrough came in 2020, when AlphaFold2 was unveiled, solving many of the most difficult protein folding problems with an accuracy previously thought unattainable. AlphaFold 2’s success marked the culmination of years of research into neural networks and machine learning, areas in which DeepMind has become a global leader. The system is trained on vast datasets of known protein structures and amino acid sequences, allowing it to generalize predictions for proteins it has never encountered—a feat that was previously unimaginable. Earlier this year, Google DeepMind and Isomorphic Labs unveiled AlphaFold 3, the third generation of the model, which