1
Google Bard/Gemini / How to Use Google Gemini - Including New Prompts
« Last post by Badshah Mamun on Yesterday at 02:14:57 PM »How to Use Google Gemini - Including New Prompts
i=aXv_ORjne703r6B0
News:
Daffodil International University Forum contains information about Open Text material, which is only intended for the significant learning purposes of the university students, faculty, other members, and the knowledge seekers of the entire world and is hoped that the offerings will aide in the distribution of reliable information and data relating to the many areas of knowledge.
Recent Posts
2
Google Bard/Gemini / Google Gemini Full Tutorial For Beginners In Hindi 2024 - 10 Tips
« Last post by Badshah Mamun on Yesterday at 02:07:40 PM »Google Gemini Full Tutorial For Beginners In Hindi 2024 - 10 Tips
i=gZyixGXZ7nWXekwC
3
AI for Professionals / 5 ChatGPT Productivity Prompts That Can Cut Your Workload In Half
« Last post by Imrul Hasan Tusher on Yesterday at 01:03:23 PM »5 ChatGPT Productivity Prompts That Can Cut Your Workload In Half
5 ChatGPT productivity prompts (cut your workload in half) GETTY
What if you could achieve the same outcome in half the time? Then you’d have choices. You could work less, or you could double up and get even further. But how do you get there? Ambitious entrepreneurs are intentional about how they spend their time. They work smart as well as hard. With so many AI tools around, there’s no excuse not to ask the question: How can I leverage what already exists to free me from my most mundane tasks?
ChatGPT can help you figure it out and make your plan. Use these prompts to do less of the mundane and create more space for your genius. Copy, paste and edit the square brackets in ChatGPT, and keep the same chat window open so the context carries through.
Work Smarter And Reduce Your Workload With ChatGPT
Stop Micromanaging
If you’ve trained someone to do something, you have to trust that they will. Otherwise, what’s the point? But leaders everywhere delegate then worry. They see something being done in a different way to how they would do it, and they panic and interfere, taking the project back over or micromanaging its new owner. Not the plan. Your job is leadership, not oversight. Shift from hands-on management to strategic leadership for the autonomy, efficiency and happiness of you and your team.
“I have recently delegated [describe task you have delegated] to [describe the role or person who now looks after it]. I’m concerned about [describe your concerns]. Act as a business coach and ask me questions, one by one, to understand whether I have good reason to be concerned, and therefore what I should do, or if I am simply being controlling and micromanaging, without any need. Ask the questions one by one, help me understand my process of delegating this task to ascertain whether I did sufficient preparation.”
Structure Your Meetings
Most Zooms could be Looms. Most Looms could be emails. Knowledge workers across the land are saying in ten sentences what could be said in three. Don’t let that be you. Start with meetings. Make every meeting matter when you remove the nonessential chatter and go in with a plan. Cut to the chase with clear agendas and defined outcomes, saving time and speeding up decision-making.
“I want to stop wasting time in meetings, and require every meeting to have a defined agenda that is stuck to by every member. Acting as an assertive productivity consultant, ask me questions about my next meeting, its attendees and purpose. From this information, create a concise agenda that can be shared before. Add a sentence I can use when sharing this agenda, to explain to attendees that going off topic cannot happen in this meeting, and we should work together to make that happen. Before beginning the questioning, require me to type, ‘Let’s end the meeting madness!’”
Create An Email Autoresponder
Not only is there meeting madness among entrepreneurs and business leaders, but email madness too. Most of your emails aren’t that important. Most are people asking things they could have googled, or stalling instead of taking action, passing the buck to buy them more time. Let’s end the email madness with an all-encompassing autoresponder that empowers someone to find the answer and frees you of instant responses. A good one of these can be a lifesaver. No one will be offended to receive it, but they’ll understand how busy you are. Help them help themselves. Manage expectations and filter urgency, keeping you focused on your real work.
“Create an email autoresponder so people who email me can progress their work without waiting for a response from me. Open a dialogue where you ask me what people often email me about, inviting me to paste in a typical message I receive. Next, ask for my typical answer. Keep going with these emails, one by one, until we have covered the most popular topics. Then, collate the information into a concise yet helpful email responder that I can put in place. At the start of the email, explain that I get a lot of enquiries and I put these FAQs together to help people find a way forward. Use my writing style to create the response: [Include an email you wrote for context]..”
Automate Your Perfectionism
Perfectionism could be holding you back. If you’re operating at a high standard, you’re going to make mistakes. Progress means learning from errors and going again, not beating yourself up or trying to avoid them all together, perhaps by not showing up at all. Use ChatGPT to help with your editing, then let good enough be your best friend. Maintain high standards without the pitfalls of perfectionism.
“I am a recovering perfectionist. Perfectionism has held me back in the past, but not any more. Your task is to assess the work I’m about to publish and find any errors that I should fix. After telling me where the errors are, tell me that you’re proud of the effort I put into this work. Help me realize that putting in maximum effort and being prepared to learn and iterate is more important than everything being exactly right the first time. Not everything requires perfection, and shipping the work is often more important.”
Organize Your Priorities
Focus fuels success. People who get distracted by breaking news, notifications and endlessly scrolling newsfeeds are neglecting their genius. Make focus your superpower. Develop a system that separates must-dos from might-dos, ensuring your energy is invested in high-impact activities. When you’re doing your work, be all in. Shut off everything else, close the door, put your phone on silent. Nothing should be able to pop up and steal you away, you’re better than that.
“Your task is to organize my day’s work in order of what will make the most difference to my business. I will paste my task list, and you should ask me questions, one by one, first to ascertain my number one business goal and then to work out which tasks contribute to it. All that matters is my one business goal, everything else can wait. Help me stick to that one goal by prioritizing my work and reminding me why it’s important to do it in that order and not get distracted.”
Cut Your Workload In Half — ChatGPT Prompts To Do Only What Matters
If you could stop micromanaging, structure every meeting and stick to the plan, reduce pressure to respond to emails, manage your perfectionism and follow your task list in priority order, you’d be unstoppable. There would be a machine-like magic to your work. People would know you were on a mission. The stars would align and the path to your dream future would appear in front of you. Most people don’t get there. They waste time, they waste chances and they waste their potential. Use these ChatGPT prompts to open your mind and uncap your limits.
Source: https://www.forbes.com/sites/jodiecook/2024/04/15/5-chatgpt-productivity-prompts-cut-your-workload-in-half/?sh=2903dcbc1b4d
4
Robotics / FLYING CAR LEADERS: ARCHER & JOBY AVIATION
« Last post by Imrul Hasan Tusher on April 15, 2024, 03:36:36 PM »FLYING CAR LEADERS: ARCHER & JOBY AVIATION
Flying cars are finally here.
You can call them “eVTOLs” (electric vertical take-off/landing) or “air taxis,” I personally will call them flying cars.
After decades of waiting, the convergence of a few key technologies and factors will enable commercial service to start in 2025:
DEP or direct electric propulsion: special electric motors
Batteries: higher energy density, cheaper batteries, mainly driven by Tesla
Materials: lightweight, strong materials
Sensors: a new generation of sensors
Computation/AI: the ability to integrate all data for safe flights
Regulatory Support: governments are finally ready to license this tech
Regarding this last bullet, in November 2022 the US Federal Aviation Administration (FAA) proposed new rules that help pave the way for commercial air taxi operations by 2025, adding something called “powered-lift” operations to its regulations.
Former acting FAA Administrator Billy Nolen has said this about the future timeline:
“We know that when the Los Angeles Olympics get underway in 2028, air taxis will be in high demand. We may see some of them in the years leading up, but nowhere near the scale in 2028.”
Industry reports suggest the potential for a $30 billion marketplace by 2030.
In today’s blog, we’ll look at the two leading flying car companies: Archer Aviation and Joby Aviation.
Let’s dive in…
Archer Aviation
Last month, I hosted Archer Aviation CEO Adam Goldstein and Chief Commercial Officer Nikhil Goel at the Abundance Summit.
In September of 2021, Archer Aviation went public (via SPAC) for $3.8 billion.
Today, their flying car design called Midnight boasts an impressive performance envelope:
Payload: Pilot + 4 paying passengers + luggage
Propulsion: 12 electric engines supported by 6 independent battery packs
Range: Up to 100 miles
Speed: Up to 150 miles per hour
Altitude: Typically, 1,500 feet (below 5,000 feet)
Charge time: 12-minute charge time between back-to-back 20-mile flights
The year 2023 marked a pivotal milestone in the development of Midnight, as the company conducted its first full-scale, uncrewed, and tethered test flight. This achievement, the result of four years of rigorous flight testing, paved the way for further advancements.
Looking ahead, Archer is poised to conduct an astounding 400 tests of its Midnight aircraft in 2024, a testament to their unwavering dedication to perfecting this groundbreaking technology.
On the regulatory front, Archer has made significant strides. The Federal Aviation Administration (FAA) has recently approved certification plans for Archer's production aircraft, and the company has announced that the first three piloted aircraft are currently under construction. These conforming Midnight aircraft will begin piloted flight testing later this year and will subsequently undergo “for credit” flight testing with the FAA as Archer progresses towards commercialization.
The company has secured an impressive indicative order book of up to 700 aircraft, valued at $3.5 billion, from major players such as United Airlines in the US, InterGlobe in India, and Air Chateau in the United Arab Emirates.
So, when can we expect to see Archer's Midnight overhead?
The company has set an ambitious goal of bringing the Midnight eVTOL to market by 2025. In partnership with Atlantic Aviation, Archer is developing electric aircraft infrastructure at existing assets, including Santa Monica Municipal Airport (SMO). Early launch markets will focus on highly congested cities such as Los Angeles, New York, and Miami, with initial routes connecting airports to city centers. As availability of the Midnight increases, services will expand to other locations across Atlantic's portfolio.
To illustrate the transformative potential of eVTOL technology, consider a trip from Santa Monica to Malibu. While this 12-mile journey could take over an hour by road, an air taxi would cover the distance in a mere five minutes, with each passenger paying roughly $30 to $40—less than the cost of a rideshare vehicle.
Joby Aviation
Founded 14 years ago, Joby was the first serious flying car company, and the first to go public in August of 2021 for $4.5 billion. And, even more impressive, in 2022, Joby distinguished itself as the first eVTOL firm to receive US airworthiness certification—a notable badge of honor in a burgeoning industry.
Among the main investors bolstering Joby's successes are Delta Air Lines and the automotive giant Toyota (which has actively aided the air taxi manufacturer in its plans to erect a factory in Ohio).
Here are the details of Joby’s eVTOL performance:
Payload: Pilot + 4 paying passengers + luggage (total capacity of 1,000 lbs.)
Propulsion: 6 electric dual-wound motors on 6 tilt-prop propellers
Range: Up to 150 miles
Speed: Up to 205 miles per hour
Altitude: Typically, 1,500 feet (below 5,000 feet)
In September 2023, Joby signed a significant contract with the US Air Force, valued at up to $131 million, and delivered its first eVTOL to Edwards Air Force Base in Southern California. This partnership involves collaboration with NASA to research the aircraft's performance in urban environments, providing valuable insights for air taxi development and the FAA.
Joby recently announced plans for a $500 million manufacturing plant in Ohio, set to begin construction in 2024, with the capacity to produce up to 500 aircraft annually.
CEO JoeBen Bevirt envisions Joby's eVTOLs as an integral part of aerial ridesharing networks by 2025. This vision took a significant step forward in November 2023 when Joby conducted the first eVTOL test flights in New York City. Furthermore, in February 2024, the company secured an exclusive six-year deal to operate air taxis in Dubai, with commercial operations expected to begin by early 2026.
The skies above are about to get a lot more interesting.
Why This Matters
Flying cars promise to redefine not just transportation, but also our very perception of accessibility and proximity.
As Archer gears up to manufacture 2,000 Midnight vehicles per year, and Joby exceeds 500 eVTOLs per year, the total production rate of flying cars will rival the production of all other flying aircraft put together.
We are poised to embrace a world where the distant becomes near, the inaccessible becomes reachable, and where time, traditionally lost in transit, is reclaimed.
Another benefit of eVTOLs will be an expansion of human connection and interaction. Communities once isolated by geographical challenges will now become integral parts of urban tapestries.
The age-old dichotomy of urban hustle and rural tranquility may very well converge, creating a harmonious blend of both worlds. As eVTOL technology continues its ascent, we are not just witnessing the evolution of travel; we are partaking in a holistic transformation of human experience.
Source: https://www.diamandis.com/blog/abundance-48-flying-car-leaders-archer-joby
5
Generative AI / AI Coding Is Going From Copilot to Autopilot
« Last post by Imrul Hasan Tusher on April 15, 2024, 02:52:26 PM »AI Coding Is Going From Copilot to Autopilot
ANDRIY ONUFRIYENKO/GETTY IMAGES
A new breed of AI-powered coding tools have emerged—and they’re claiming to be more autonomous versions of earlier assistants like GitHub Copilot, Amazon CodeWhisperer, and Tabnine.
One such new entrant, Devin AI, has been dubbed an “AI software engineer” by its maker, applied AI lab Cognition. According to Cognition, Devin can perform all these tasks unassisted: build a website from scratch and deploy it, find and fix bugs in codebases, and even train and fine-tune its own large language model.
Following its launch, open-source alternatives to Devin have cropped up, including Devika and OpenDevin. Meanwhile makers of established assistants have not been standing still. Researchers at Microsoft, GitHub Copilot’s developer, recently uploaded a paper to the arXiv preprint server introducing AutoDev, which uses autonomous AI agents to generate code and test cases, run tests and check the results, and fix bugs within the test cases.
“It’s exciting to see more versions of AI coding assistants with new capabilities,” says Ben Dechrai, a coder and developer advocate at software company Sonar. “They validate the need for generative AI tools in developers’ workflows.”
Dechrai adds that these coding copilots can help software engineers write code faster, allowing them to focus on more strategic and creative tasks. Another advantage of these programming tools is the ability to create a template for code, notes Saurabh Bagchi, a professor of electrical and computer engineering at Purdue University. Much as with prompt engineering, developers must provide these assistants with “the right kind of software requirements to produce a template, and then a software engineer can fill in the gaps,” he says.
These gaps include safety and reliability considerations. Software engineers must look out for security vulnerabilities in AI-generated code, as well as the types of corner cases that could cause it to crash.
“Developers still need to ensure rigorous quality standards are in place when analyzing and reviewing code written with generative AI, just as they would with code developed by a human,” says Dechrai. “AI coding assistants are good at suggesting code, reflecting on the code, and reasoning about its effectiveness, but even then it’s not 100 percent accurate.”
Dechrai cautions that autonomous coders are “still so new that developers are just learning which use cases will be most beneficial.” And they’ll need to be “ironed out in the real world to see how much they’re able to deliver on their promise,” says Bagchi.
AI Coders vs. the Humans
Doom-and-gloom predictions of replacing human software engineers are also bound to follow the emergence of these “AI software engineers,” but that won’t be happening anytime soon. Devin, for instance, resolved only 14 percent of a subset of GitHub issues from real-world code repositories. “There’s still a long way to go for it to become something I can rely on blindfolded,” says Bagchi.
He notes that these autonomous programming tools have another blind spot: the fact that software development happens in collaboration. Coding copilots try to do everything, and they might do it reasonably well. On the other hand, different software engineers have their own specialties—be it front end, back end, full stack, or data, to name a few—and they all work together to build a cohesive product.
“To develop intuitive systems, you need an iterative process with humans in the loop to provide feedback,” Bagchi says. “The fundamental human intuition, depth, and imagination has to be brought to bear.”
That’s why Bagchi believes these unassisted versions won’t be dominating the space that coding assistants hold—at least for now. “The models running underneath are similar in architecture, and as technology continues to evolve, both of them will get better,” he says. “But the Copilot or CodeWhisperer model seems most promising and is better suited to complex software development where humans work with the assistance of AI.”
Yet programmers “should start using these tools if they haven’t already, or they’ll risk getting left behind,” says Dechrai. “If you want to know if an AI coding assistant is truly beneficial, you have to use it yourself, get to know it, and see where it fails.”
Bagchi echoes the sentiment: “Try them out with the use cases you have and stress them with the kinds of software you’re creating.” But because unassisted coding copilots are a nascent technology, they are likely to improve rapidly. “So you have to track them,” he adds.
Moreover, software engineers will have to “consistently ensure code is secure, reliable, and maintainable throughout its life cycle,” Dechrai says. “It will always be up to the developer to properly understand the output and how it was generated.”
Source: https://spectrum.ieee.org/ai-code-generator
6
Robotics / RECOVERING 216 HOURS PER YEAR: TESLA & WAYMO
« Last post by Imrul Hasan Tusher on April 15, 2024, 02:47:19 PM »RECOVERING 216 HOURS PER YEAR: TESLA & WAYMO
Autonomous, electric, ride-sharing “car-as-a-service” (using robotaxis) has the potential to be 80% cheaper than individual car ownership.
The average U.S. commuter currently spends a mind-numbing 52 minutes each day roundtrip, trapped in the confines of their vehicle—for average employees that’s equivalent to recovering 216 hours, or twenty-seven 8-hour days, during your year.
Freed from the task of driving, this time could be transformed into periods of relaxation, productivity, or even leisure. Ever thought of finishing a novel, practicing a new language, or perhaps indulging in some personal passions during your commute?
The era of the robo-chauffeur promises just that.
In today’s blog, I’ll discuss the top two companies leading this transportation revolution... Tesla & Waymo.
Let’s dive in…
Tesla
Tesla has a secret plan is to deliver a global fleet of robotaxis—and I have every confidence Elon will make that happen. What’s his secret weapon? The release of Tesla’s newest iteration of Full Self-Driving or FSD 12, a radical departure from Tesla's previous versions.
Unlike prior systems, FSD 12 wasn't created from meticulously crafted lines of C++ code. Instead, it arose from the technology's ability to learn autonomously, imbibing billions of video frames to mimic human drivers.
A Tesla engineer, Dhaval Shroff, likened it to ChatGPT, stating, "It's like ChatGPT, but for cars."=
The essence here was replicating human learning processes through neural networks, processing massive data volumes to simulate human actions in intricate driving scenarios.
Historically, Tesla's Autopilot used a rules-based approach, a structure where specific situations triggered codified reactions. Lane markings, pedestrian movements, vehicles, signals, each elicited a programmed response. But FSD 12 turned this approach on its head. Instead of following rigid rules, Shroff's "neural network planner" emulated human behavior, learning not from pre-set conditions but from observing human drivers' actions in real-life scenarios.
Central to Tesla’s robotaxi objectives is a singular metric—"miles driven without human intervention.”
Elon's directive was clear: "I want the latest data on miles per intervention to be the starting slide at each of our meetings."
The goal?
Push the limits of the neural network until it surpassed human driving capabilities.
The team's discovery that optimal performance required training on at least 1 million video clips underscored Tesla's unique advantage in this race to full autonomy. And the way in which Tesla’s Autopilot learned is something that surprised even Elon. As he put it, “I mean the really wild thing about the end-to-end training is that it learns to read. It can read signs, but we never taught it to read … we never taught it what a car was or what a person was or a cyclist. It learnt what all those things are, what all the objects are on the road from video, just from watching videos. Just like humans.”
With nearly 2 million Tesla cars globally collecting data daily, the company was uniquely poised to train and constantly improve FSD 12.
As of April 2024, Tesla has rolled out v12.3.3 of FSD, now called FSD (Supervised), to over 5,000 cars. Under the driver’s supervision, FSD vastly improves the car’s autonomous capabilities—from making lane changes and navigating around other cars or objects, to making left and right turns and parallel parking. (As of the writing of this blog, Tesla is already beginning its initial rollout out of FSD (Supervised) v12.3.4.)
I personally love my Model S on FSD... 99.9% of time it gets me all the way to my destination, and it’s only tried to kill me once!
Waymo
Beginning life as Google's internal experiment in 2009, Waymo evolved under Alphabet into a dedicated venture aimed at bridging the vast divide between self-driving vehicles and revenue generating robotaxis.
In its commitment to reshaping the future, Waymo vehicles have traversed over 12 million miles since 2009, both in real-world environments and simulations.
The company's profound objective is to eradicate human errors that account for countless fatalities annually. Leveraging state-of-the-art camera and Lidar laser technology, Waymo's fleet is equipped to visualize the world in incredible detail, irrespective of the hour. This ability to navigate with precision owes much to real-time sensor data fused with intricately detailed custom maps. And with the computational prowess of server-grade GPUs and CPUs, Waymo's onboard systems can process this flood of data, ensuring that passengers experience a journey that's not only safe but efficiently plotted.
In August 2023, Waymo made history in San Francisco, as the company began allowing the public to pay for rides in its driverless cars.
This groundbreaking development saw Waymo's vehicles functioning as true robotaxis, offering a glimpse into a future where the traditional taxi experience is reimagined. With the Waymo One app, booking a ride is as intuitive as hailing an Uber, but what arrives is a gleaming white Jaguar from Waymo's 250-strong fleet. These vehicles, each valued at a staggering $200,000, are outfitted with an array of high-tech sensors and cameras, ensuring an unparalleled level of safety and efficiency.
Yet, despite their technological sophistication, rides remain affordable, with fares ranging from $18 to $21, on par with traditional ride-hailing services.
As Waymo continues to push the boundaries of autonomous transportation, the company's recent expansion into Los Angeles marks a significant milestone.
In March 2024, Waymo secured approval from the California Public Utilities Commission (CPUC) to extend its services to select areas of Los Angeles and the Bay Area. With an initial fleet of fewer than 50 cars, Waymo's operational territory encompasses approximately 63 square miles, stretching from the coastal charm of Santa Monica to the vibrant heart of downtown Los Angeles.
While the service currently excludes airport trips and freeway travel, the demand is palpable, with a waitlist of 50,000 eager Angelenos ready to experience the magic of autonomous rides.
For Waymo, the Los Angeles expansion represents more than just another market—it is a proving ground for the company's vision of the future. With a metropolitan population of 13 million, LA's intricate web of freeways, narrow streets, and unprotected left turns, coupled with its notorious traffic and distracted drivers, poses a formidable challenge.
Yet, the potential rewards are immense, with estimates suggesting that the LA market could generate up to $2 billion in revenue for the company.
Why This Matters
With the rise of autonomous vehicles, the landscape of the auto industry is poised for a seismic shift.
Currently home to over a hundred brands, the next decade will witness a substantial consolidation in the automotive realm.
Two main forces will drive this: car usage rates and functionality.
Most cars today are utilized less than 5% of their potential, often gathering dust in driveways. With the advent of car-as-a-service, fewer vehicles will serve more people, disrupting the demand-supply chain.
Additionally, in this new market, brand loyalty will wane. Consumers, attracted by efficiency and cost-effectiveness, will care less about the brand and more about the service.
Hence, a significant reduction in the number of automaker brands is anticipated, challenging giants in Detroit, Germany, and Japan.
Donald Shoup, an esteemed urban planning professor at UCLA, provides insight into another profound implication: the impact on real estate. With a staggering 2 billion parking spots in the U.S., Shoup points out the startling fact that “the area of parking per car in the United States is thus larger than the area of housing per human.”
Furthermore, he reveals the hidden costs of “free parking,” estimating U.S. expenditures between $102 billion to $374 billion—somewhere in the ballpark of the Medicare and national defense budgets. But what if these vast parking spaces become redundant? With autonomous vehicles on-demand, the demand for parking diminishes. Our cities could witness a commercial real estate boom, or perhaps, some of these spaces might metamorphose into thriving community centers or lush green parks.
The future of autonomous vehicles isn't merely about technological progression—it's about reshaping societies, economies, and urban landscapes.
But these aren’t the only vehicles transforming mobility. As we’ll see in the next few blogs, flying cars are finally on their way.
Source: https://www.diamandis.com/blog/abundance-47-dawn-of-the-robo-chauffeur
7
Generative AI / Why small language models are the next big thing in AI
« Last post by Imrul Hasan Tusher on April 15, 2024, 02:41:16 PM »Why small language models are the next big thing in AI
In the AI wars, where tech giants have been racing to build ever-larger language models, a surprising new trend is emerging: small is the new big. As progress in large language models (LLMs) shows some signs of plateauing, researchers and developers are increasingly turning their attention to small language models (SLMs). These compact, efficient and highly adaptable AI models are challenging the notion that bigger is always better, promising to change the way we approach AI development.
Are LLMs starting to plateau?
Recent performance comparisons published by Vellum and HuggingFace suggest that the performance gap between LLMs is quickly narrowing. This trend is particularly evident in specific tasks like multi-choice questions, reasoning and math problems, where the performance differences between the top models are minimal. For instance, in multi-choice questions, Claude 3 Opus, GPT-4 and Gemini Ultra all score above 83%, while in reasoning tasks, Claude 3 Opus, GPT-4, and Gemini 1.5 Pro exceed 92% accuracy.
Interestingly, even smaller models like Mixtral 8x7B and Llama 2 – 70B are showing promising results in certain areas, such as reasoning and multi-choice questions, where they outperform some of their larger counterparts. This suggests that the size of the model may not be the sole determining factor in performance and that other aspects like architecture, training data, and fine-tuning techniques could play a significant role.
The latest research papers announcing new LLMs all point in the same direction: “If you just look empirically, the last dozen or so articles that come out, they’re kind of all in the same general territory as GPT-4,” says Gary Marcus, the former head of Uber AI and author of “Rebooting AI,” a book about building trustworthy AI. Marcus spoke with VentureBeat on Thursday.
“Some of them are a little better than GPT-4, but there’s no quantum leap. I think everybody would say that GPT-4 is a quantum step ahead of GPT-3.5. There hasn’t been any [quantum leap] in over a year,” said Marcus.
As the performance gap continues to close and more models demonstrate competitive results, it raises the question of whether LLMs are indeed starting to plateau. If this trend persists, it could have significant implications for the future development and deployment of language models, potentially shifting the focus from simply increasing model size to exploring more efficient and specialized architectures.
Drawbacks of the LLM approach
The LLMs, while undeniably powerful, come with significant drawbacks. Firstly, training LLMs requires an enormous amount of data, requiring billions or even trillions of parameters. This makes the training process extremely resource-intensive, and the computational power and energy consumption required to train and run LLMs are staggering. This leads to high costs, making it difficult for smaller organizations or individuals to engage in core LLM development. At an MIT event last year, OpenAI CEO Sam Altman stated the cost of training GPT-4 was at least $100M.
The complexity of tools and techniques required to work with LLMs also presents a steep learning curve for developers, further limiting accessibility. There is a long cycle time for developers, from training to building and deploying models, which slows down development and experimentation. A recent paper from the University of Cambridge shows companies can spend 90 days or longer deploying a single machine learning (ML) model.
Another benefit of SLMs is their potential for enhanced privacy and security. With a smaller codebase and simpler architecture, SLMs are easier to audit and less likely to have unintended vulnerabilities. This makes them attractive for applications that handle sensitive data, such as in healthcare or finance, where data breaches could have severe consequences. Additionally, the reduced computational requirements of SLMs make them more feasible to run locally on devices or on-premises servers, rather than relying on cloud infrastructure. This local processing can further improve data security and reduce the risk of exposure during data transfer.
SLMs are also less prone to undetected hallucinations within their specific domain compared to LLMs. SLMs are typically trained on a narrower and more targeted dataset that is specific to their intended domain or application, which helps the model learn the patterns, vocabulary and information that are most relevant to its task. This focus reduces the likelihood of generating irrelevant, unexpected or inconsistent outputs. With fewer parameters and a more streamlined architecture, SLMs are less prone to capturing and amplifying noise or errors in the training data.
Clem Delangue, CEO of the AI startup HuggingFace, suggested that up to 99% of use cases could be addressed using SLMs, and predicted 2024 will be the year of the SLM. HuggingFace, whose platform enables developers to build, train and deploy machine learning models, announced a strategic partnership with Google earlier this year. The companies have subsequently integrated HuggingFace into Google’s Vertex AI, allowing developers to quickly deploy thousands of models through the Google Vertex Model Garden.
After initially forfeiting their advantage in LLMs to OpenAI, Google is aggressively pursuing the SLM opportunity. Back in February, Google introduced Gemma, a new series of small language models designed to be more efficient and user-friendly. Like other SLMs, Gemma models can run on various everyday devices, like smartphones, tablets or laptops, without needing special hardware or extensive optimization.
Since the release of Gemma, the trained models have had more than 400,000 downloads last month on HuggingFace, and already a few exciting projects are emerging. For example, Cerule is a powerful image and language model that combines Gemma 2B with Google’s SigLIP, trained on a massive dataset of images and text. Cerule leverages highly efficient data selection techniques, which suggests it can achieve high performance without requiring an extensive amount of data or computation. This means Cerule might be well-suited for emerging edge computing use cases.
Another example is CodeGemma, a specialized version of Gemma focused on coding and mathematical reasoning. CodeGemma offers three different models tailored for various coding-related activities, making advanced coding tools more accessible and efficient for developers.
The transformative potential of small language models
As the AI community continues to explore the potential of small language models, the advantages of faster development cycles, improved efficiency, and the ability to tailor models to specific needs become increasingly apparent. SLMs are poised to democratize AI access and drive innovation across industries by enabling cost-effective and targeted solutions. The deployment of SLMs at the edge opens up new possibilities for real-time, personalized, and secure applications in various sectors, such as finance, entertainment, automotive systems, education, e-commerce and healthcare.
By processing data locally and reducing reliance on cloud infrastructure, edge computing with SLMs enables faster response times, improved data privacy, and enhanced user experiences. This decentralized approach to AI has the potential to transform the way businesses and consumers interact with technology, creating more personalized and intuitive experiences in the real world. As LLMs face challenges related to computational resources and potentially hit performance plateaus, the rise of SLMs promises to keep the AI ecosystem evolving at an impressive pace.
Source: https://venturebeat.com/ai/why-small-language-models-are-the-next-big-thing-in-ai/
8
Common Forum/Request/Suggestions / Outstanding Сasual Dating - Actual Women
« Last post by ruhul5710 on April 13, 2024, 08:29:03 AM »9
Common Forum/Request/Suggestions / Prime Сasual Dating - True Females
« Last post by ankhi739 on April 12, 2024, 11:22:21 PM »10
Robotics / THE BIRTH OF ROBOTAXIS (CAR-AS-A-SERVICE)
« Last post by Imrul Hasan Tusher on April 11, 2024, 09:54:42 PM »THE BIRTH OF ROBOTAXIS (CAR-AS-A-SERVICE)
How would your life improve if you never had to drive again? If traveling across town to work or your friend’s home was cheap and fully autonomous, enabled by a robotic chauffeur?
How much time would you save if you weren’t the one behind the wheel? What would you do with those extra hours?
Over the next few blogs in this Age of Abundance series, we’ll discuss an additional category of advanced robotics, namely autonomous vehicles, flying cars or eVTOLs (electric Vertical Take-off or Landing), and delivery robots helping to get people and goods from one point to another.
Fully autonomous vehicles from Tesla and Waymo (to name a few) are on the path to enable “car-as-a-service” fleets (or robotaxis) operating on-demand, Uber-like services.
The cost of ground transportation is slated to decrease between 2x to 4x as a result. Sometime in the near future, your kids or elderly parents will never drive.
A significant percentage of parking garages, driveways, and parking structures will eventually be transformed into alternative usable space. Autonomous cars will take all shapes and sizes and serve as functional “third spaces” used for entertainment, sleeping, or meeting rooms as drive time becomes work or play time.
Meanwhile, aerial ridesharing, eVTOLs, and flying cars will also become fully operational in most major metropolitan cities this next decade.
Where you live and work will begin to transform as these systems shrink travel time and distance. Previously difficult to reach geographies (islands, rural areas, mountain tops) will become accessible.
Individuals seeking the solitude of the country will also have access to the shopping, food, and entertainment of metropolitan city centers, connected through eVTOL technology.
In today’s blog, we’ll go back in time to the early 2000s, when a series of Grand Challenges laid the foundation for today’s autonomous vehicle industry.
Source: https://www.diamandis.com/blog/abundance-46-birth-of-robotaxis