Waze Conversational Reporting: Just tell Gemini AI what you see

All this of course raises critical questions about the sustainability of generative AI and about our own carbon footprints. The AI companies themselves are reluctant to tell us exactly how much energy they use, but they apparently can’t stop their own chatbots having a stab. I asked ChatGPT-4 “how much energy was used to process this query? ” and it said “0.002 to 0.02 kWh”, which it said “would be similar to keeping a 60-watt bulb on for about 2 minutes”. Unsurprisingly, the more powerful the AI, the more energy it consumes.

The company said it’ll launch in preview in December as part of its UiPath Studio developer tool suite. It will give developers everything they need to design, build, evaluate and publish AI-powered agents that can collaborate with its traditional process automation robots. Last week, OpenAI got into the search engine business with its generative AI-powered ChatGPT Search. This search engine provides detailed answers to questions entered into a search bar, drawn from the information in its generative AI model.

How foreign operations are manipulating social media to influence your views

It’s a bit more about improving and augmenting what we’ve got than layering on more and more people. Forbes senior contributor John Koetsier did some head-to-head tests of Google and ChatGPT Search to see which search engine gave the most accurate and informative result. Of the 10 queries, ChatGPT Search answered four better, Google answered three better, and three were a tie. ChatGPT can provide better detailed information—like figuring out which bidet is the best to buy—while Google calls upon more credible sources. And while Google has decades of web crawling to inform its findings, Koetsier points out that it also serves up lots of ads, which sometimes detract from the results. Another critical component of agentic AI is its deep reasoning capabilities.

We want our readers to share their views and exchange ideas and facts in a safe space. As for the first steps when using agentic AI, you should ensure you already have clean data.

Moreover, it can assist in generating high-quality content and powering chatbots. However, despite these tangible benefits, GenAI lacks the ability to take action on behalf of the users. Its functionality remains limited in scope and is prone to hallucinations. Interestingly, the definition of agentic AI remains fuzzy, as the technology is still in its nascent stages.

OpenAI taking on Google Search with prototype of SearchGPT

This can be expensive and risky, requiring sophisticated infrastructure and talented data scientists. However, as mentioned above, it must have an AI-native architecture and many integrations with enterprise systems and applications. I recommend choosing a vendor with many customers that have demonstrated clear ROI.

It can then set up the employee’s profile, assign benefits and enroll them in payroll. Simultaneously, the AI can integrate with IT systems to create email accounts, set permissions and configure access to necessary applications and platforms. To start, I recommend building agentic AI on an AI-native architecture as a fundamental step that can help future-proof in a rapidly evolving tech landscape. Seamless integration with modern AI frameworks, automation and orchestration tools are also critical. Without these, you risk ending up with a standard GenAI solution lacking the autonomy, depth and versatility that true agentic AI delivers.

Folks who might be mid-career who are struggling to change how they work, I think, are generally under threat. I believe agentic AI offers a transformative opportunity for enterprises that can go beyond the limitations of GenAI. Its core characteristics—autonomy, deep reasoning, reinforced learning and integration with tools—can help you initiate, execute and optimize complex workflows with minimal human intervention. Salesforce’s Agentforce, for example, provides AI-powered conversational agents for CRM, marketing and data management. CEO Marc Benioff even predicts there will be one billion AI agents by 2026.

GenAI large language models (LLMs) lack the ability to perform complex reasoning or take direct actions, which can greatly diminish their potential productivity gains. And quite frankly, these foundational LLMs can be prohibitively expensive to deploy in an enterprise environment. These systems also excel at reasoning and making complex decisions based on context, employing reinforcement learning to adapt through interaction with their environment. Google is focusing on creating generative AI that changes and enhances the face of commonly used tools to be readily available, productive, and creative.

Google’s Waze app is so popular with drivers because of its unique incident reporting feature, which helped it stand out from the crowd of navigation apps many years ago. Since then, Google has continued to improve Waze, and it leveled the playing field a bit by bringing support for incident reporting to Google Maps. Third, consider choosing less energy-demanding social media, using environmental ranking information to inform the decision. Reducing the amount of time spent on social media can directly decrease energy consumption. Nvidia surpassed Apple to hit that mark on Monday morning, and has largely stayed there. It hit an all-time high on Wednesday, with its market cap at $3.57 trillion, and its share price reaching an all-time high of $146.49.

Examples include the “Help Me Write” tool in Google Docs, which depends on generative AI to provide users with a draft e-mail, report, or any other document that entails text. It also has the effect of saving people time and conquering writer’s block. The “Magic Fill” of Google Sheets also forms a pattern in data analysis. Google revealed statistics that those who applied the AI-based tools stand a 30% chance of completing their jobs on time. It is a definite estimate that more than 90% of people begin with some kind of search query, and this behavior just increases with AI-powered options for search.

We’ve talked about this for years, aligning technology and business, but it truly is happening now in a business context. You could be providing 500 call center workers for a global company. That client might come back and say, I don’t want to keep spending $10 million a year. What you need to do is think, how do I scale my business with these same 500 people, without layering on more and more staff all the time as it scales.

Your job may be under threat from somebody else who has better AI competency than you. So if someone’s more AI-literate, they know how to operate these models, there’s a high chance your company is looking at you and thinking, is this person evolving with the times? You talk to a lot of these Gen Z kids coming out of college now. They’re familiar working in these environments and learning how to use these new technologies.

I talked to Phil Fersht, CEO of HFS Research, about how the move toward agentic AI is impacting businesses. Traditional intent-based systems are a current hurdle because they sometimes misinterpret user queries if the exact intent isn’t defined. Agentic AI, however, can help act on complex requests, delivering a more intuitive conversational experience that can accelerate decision-making and enhance user satisfaction. It has proven highly effective at generating software code and enhancing content management through enterprise search or RAG.

Generative AI lets Google better understand and respond to complex, conversational search queries, providing a more accurate and intuitive search experience. The technology can now better interpret the natural language inputs that may provide a more personalized response than mere links on the web. This way, users receive more information about answers, summaries, and insights on even the most niche queries. AI systems operate on a query response basis without maintaining long-term context.

As businesses continue to navigate an evolving technological landscape, I encourage you to test how agentic AI can help you deliver enterprise value. Of course, just like with past AI applications, agentic AI systems should be built on rigorous ethical frameworks, with secure design and deployment practices ChatGPT to mitigate potential risks. One challenge to avoid is the proliferation of standalone SaaS app-based AI copilots. Instead, there should be a unified interface that is accessible anywhere, whether in your email, Slack or mobile app. This means having an enterprise-wide, universal and agentic AI copilot.

Generative AI is energy-hungry

Sign up for the most interesting tech & entertainment news out there. On a train or bus, or just standing in a queue, the most common sight these days is the muted glow of a screen, and the flickering thumbs of people lost in the endless scroll on their smartphones. Our community is about connecting people through open and thoughtful conversations.

Generally, agentic AI refers to systems characterized by autonomy. They can autonomously initiate and complete tasks, making real-time decisions and dynamically taking actions with minimal to no human supervision. Every time we read an article, see an advertisement, watch a photo or video, that generative ai vs conversational ai content needs to be transferred from the social media platform’s servers to our device. And from there, the supply chain continues, getting that product delivered to the store, warehouse or end consumer. An agentic AI system could generate and send required documents for digital signatures.

It estimates that at least 30% of GenAI projects will fail to progress beyond the proof-of-concept stage by the end of 2025. Key factors include poor data quality, inadequate risk controls, higher costs and unclear business objectives. The starting set includes plugins from Expedia, FiscalNote, Instacart, KAYAK, Klarna, Milo, OpenTable, Shopify, Slack, Speak, and Zapier, as well as the Wolfram plugin mentioned above. But when it’s connected to the Wolfram plugin it can do these things,” Wolfram wrote in a blog post. GitHub Copilot X utilizes the new GPT-4 model and is a major upgrade to the Copilot product, adding new areas where it can be used and introducing chat and voice capabilities. Altman also explained that ChatGPT conversation history was unavailable from 4 AM EST on March 22 to 1 PM that same day as they fixed the issue.

What’s new in generative AI: GPT-4 ChatGPT conversation history bug ChatGPT plugins

On the other hand, AI agents are better at adapting to new challenges, making intelligent decisions and handling complex, multistep processes. While the environmental impact of these technologies raises valid concerns, it’s also essential to recognise their benefits. To take one example, AI-assisted tools like text-to-speech, voice recognition and auto-captioning have already made society more inclusive particularly for disabled or neurodiverse people. I don’t want to suggest we scrap social media or reject generative AI entirely. What I do think is going to happen is there’s less and less need for transactional roles, and more and more need for context-filled roles in companies.

The creator Stephen Wolfram first talked about the possibility of connecting the two technologies back in January, and the two companies have been working together since to make it happen. Simply scrolling through the app exchanges a lot of data as Tiktok is constantly running videos, including many preloaded in the background that you may never even see. Moving data across the internet requires energy, sending signals through various electronic devices, including routers, servers, and our own mobile phone or laptop.

According to Google’s latest report, adaptation of Bard has been happening fast, with thousands interacting daily. Copilot is also being integrated into documentation, with a chat interface that will allow developers to ask questions about the languages, frameworks, and technologies that their code is using. It has already created this functionality for React, Azure Docs, and MDN documentation, and plans to also bring this to internal documentation as well. Another part of Copilot X is that it will be able to generate descriptions of tags and descriptions for pull requests.

This conversation had been edited for length, clarity and continuity. UiPath made its name in the area of robotic process automation, which is a subset of AI that’s focused on machine automation. Its platform provides tools that enterprises can use to automate repetitive business tasks such as data entry. They work by studying how human workers complete these tasks, so they can replicate the process, freeing up those employees to carry out higher-level work. You can foun additiona information about ai customer service and artificial intelligence and NLP. By combining AI agents with its robots, UiPath says, it will enable the automation of more complex tasks that were previously impossible for anyone but humans to perform. They’ll also be able to make intelligent decisions on behalf of users, the company said.

About 86% of companies have seen budgets for third-party risk management increase in the last year. But only 32% of companies regularly monitor their third-party vendors, and half aren’t able to assess all of them due to challenges in resources, technology and expertise. But, the report said, more frequent monitoring is ramping up, ChatGPT App which could put a larger lock on supply chain security. Each link in the chain represents another entity taking control of that good—and another vulnerability to cyber attacks. A study from cyber defense company BlueVoyant found that 81% of organizations reported negative impacts from breaches somewhere along the supply chain.

Anything that touches the customer is very sensitive, and anything that touches your employees is very sensitive. That sort of thing has a security issue around it, which plays into peoples’ SOC 2 compliance. This isn’t a one-size-fits-all approach; each system should be customized with domain-specific LLMs grounded to enterprise data, whether in finance, IT, HR or customer service. The result can be highly accurate responses, minimized hallucinations and increased relevance—all delivered at a substantially lower cost compared to generic GenAI models. Agentic AI has surged in popularity over the past few months, with major tech companies announcing new platforms based on it.

Each of these devices consumes energy to function, while servers need to be kept cool. That data is distributed across many “server farms” (typically housed in a large warehouse with thousands of computers) around the world. If you load a video from Youtube you don’t connect to a single “Youtube data HQ” somewhere in California, but will instead gather data from many different servers often in different countries or continents. Here are some tips from YouTube personality Doctor Mike—family medicine physician Dr. Mike Varshavski—about using social media as a communication tool. The new RSA ID IQ report asked more than 2,000 cybersecurity and tech professionals about their use of passwords at work. Nvidia doesn’t report earnings until later this month, and its rally was driven by two large events.

As the CEO of a company that developed agentic AI applications before they became a hot industry trend, I know how complex this technology is to build and implement. While I am certain this is the next wave of innovation, I also understand that enterprises need to take a thoughtful approach. Meanwhile, Oracle has developed over 50 role-based AI agents for its Cloud Fusion Applications Suite, covering enterprise resource planning, human capital management, supply-chain management and customer experience. He explained that AI agents can leverage the millions of automation developed by UiPath’s customers to integrate with thousands of enterprise applications. At the same time, those agents will adhere to the strict governance controls provided by UiPath’s platform. At UiPath Forward, the startup explained, its robots are best suited for carrying out repetitive, rule-based tasks in order to improve business efficiency and reduce manual effort.

• Google revealed statistics that those who applied the AI-based tools stand a 30% chance of completing their jobs on time.
• After all, the biggest challenge companies now face is keeping systems operating smoothly—”a major step in the right direction,” Molinoff said.
• I see additional critical challenges that were not addressed by the Gartner survey.
• Third, consider choosing less energy-demanding social media, using environmental ranking information to inform the decision.

And yes, that’s a lot, but it’s a marked improvement from the 94% of companies reporting problems with these kinds of breaches last year. The reduction, BlueVoyant Global Head of Supply Chain Defense Joel Molinoff said, may come from greater awareness of supply chain risks. The UiPath Autopilot provides a conversational interface that makes it simple for any employee to take advantage of the company’s agents and workflow automations. They’ll be able to use it to find answers to their questions, grounded in the company’s own data, analyze documents, automate copy pasting across applications and more. International Data Corp. analyst Maureen Fleming said agentic automation is about the convergence of AI and rule-based automation technologies.

And then you literally have about two-thirds of organizations doing mostly nothing. You are going to see a small percentage of CIOs becoming incredibly successful at running this. If it’s a technology solution, it’s going to fail like everything else. Your ability to know that, embrace that, understand that and work with that could mean you are safe.

Last Friday, S&P Global announced that Nvidia will replace Intel in the Dow Jones Industrial Average. And on Wednesday, its stock rode a 4% bump following Donald Trump’s re-election. I think the hype around agentic AI is real, but realizing its full potential to drive ROI will demand a clear, focused strategy.

But though it is appealing, and sometimes a necessity, it comes with an environmental price tag. Another organization is saying that we need to rethink how this role of the CIO is operating. They need to straddle both business and technology, and I do think that role is going to be changed beyond our recognition in the next couple of years. Only 5% of companies today are operating at any type of scale with gen AI, and about 27% are at fairly advanced stages of piloting testing.

Better still, customers will be able to choose from a range of large language models under the hood. One of the first available is Anthropic PBC’s Claude 3.5 Sonnet, which integrates with Autopilot as well as other UiPath products, including Clipboard Ai and a new medical record summarization tool announced today. With the UiPath Agent Builder, developers will be able to build AI agents that can incorporate its RPA bots to automate various advanced business processes. They’ll be able to access a number of prebuilt agents in the UiPath Agent Catalog, or build their own from scratch, and they can also choose to integrate third-party AI agents.

The company is also working on a feature where it will warn developers if a pull request doesn’t have sufficient testing and then suggest potential tests. The new chat capabilities are intended to provide a “ChatGPT-like experience” in the editor. If you’re unsure of what you see and Gemini can’t figure out the kind of incident you’re reporting, it’ll ask you follow-up questions to get clarification before submitting the incident report on your behalf. Once enabled, you’ll just tell Waze what you see ahead, and Gemini AI will understand the type of road incident you’re reporting. You might say something like, “Looks like there are cars jammed up ahead.” Thanks to Gemini, Waze will understand that you’re reporting traffic congestion ahead, and it will submit the report.

Unlike when you stream video or load a large web page, with generative AI most energy is used at their end, while processing your query. If you ask ChatGPT to write you a novel, the process of writing involves lots of calculations, even if the resulting text itself doesn’t use much data. Generative AI, with its ability to create text, images, music and even videos, is completely reshaping lots of creative processes.

Imagine leveraging LLMs through multi-agent systems, where these specialized agents collaborate to accomplish tasks, ensuring instructions are understood and autonomously executed. An ideal agentic AI system should be vendor-agnostic and capable of connecting to hundreds of enterprise systems and applications. It must also be able to take action across the entire organization rather than being confined to a single domain to help unlock cross-functional productivity and drive meaningful impact across departments. I see additional critical challenges that were not addressed by the Gartner survey.