Most applications are built to rely on the cloud. However, many organizations are beginning to realize that cloud dependency has drawbacks. The challenges of cloud over-dependency are most apparent in industries such as aviation, quick-service restaurants, e-commerce, and healthcare, where deskless workforces rely on mobile apps to do their jobs.
Max Alexander
Using the cloud in practice
Think about how a modern fast-food restaurant uses mobile technology today. Customers input their menu choices at kiosks in the restaurant, on handhelds at the drive-thru, or at point-of-sale terminals. Their order is then sent to screens in the kitchen for the order to be made up. When it is ready, an alert is sent to counter staff to hand it over. Whilst it sounds simple, this is actually a complex, business-critical workflow in which it is vital that data is synchronized and shared in real-time.
This data must travel to a remote data center halfway around the world just to arrive at a device in the same building. Business apps become unusable if the data connection is slow, a Wi-Fi router breaks, or there is a cloud service outage.
Since cloud-only applications have so many single points of failure, operations halt whilst waiting for a device to reconnect. Apps like these do not provide a great user experience and ultimately cost businesses money, customer experience, or worse. In situations like healthcare, where decisions about a patient’s treatment are logged and updated on mobile devices used by a team, these interruptions and their impact on quick decision-making may have serious implications.
Finding ways to connect those different devices directly is an obvious solution. Indeed, peer-to-peer networking between devices is common; how many times have you used AirDrop to send pictures to a friend’s smartphone?
So why are we not more collaborative work apps built with cloud-optional features to enable them to function without the Internet?
One answer is that it has been easier to build cloud-only applications. A wealth of cloud-only databases and tools has meant developers have not had to worry about TCP/IP networking, database partitioning, or on-disk compression whenever they must update a field in a database table.
Sign up to the TechRadar Pro newsletter to get all the top news, opinion, features and guidance your business needs to succeed!
Moving beyond cloud-only applications
By contrast, offline-first applications have been challenging to implement because of the limited tools available to support them. There are scalable peer-to-peer protocols for data that do not change, like AirDropping photos. However, developers need the reliability of a database when data updates must be in real-time or super-accurate. To do this, an offline-first datastore capable of handling offline and peer-to-peer changes must be built.
But where to start? A model for a datastore that allows these in-the-field apps to work without a cloud connection has several key characteristics.
Firstly, it must be user-friendly for software developers. Instead of sending data to a remote server, the application needs to write data to its local database first in the form of changes, then detect changes from other devices and recombine them whenever needed. Devices must be able to discover, connect, and maintain these connections with nearby devices in what is called a mesh network. It is important to note that creating mesh networks for peer-to-peer data synchronization is not a silver bullet for enabling an organization to go cloud-optional when it needs to. These networks generate vast amounts of data that can overwhelm small devices if each node aggressively tries to sync every piece of data, which becomes a hindrance to businesses and halts operations.
As such, to overcome this, different types of devices should have different responsibilities when it comes to synchronizing the data. Smaller devices such as phones and tablets that have less processing power and storage capacity can focus on synchronizing the data explicitly requested by other devices, not complete documents, so as not to overwhelm the device’s bandwidth. This means that these smaller devices can sync data incredibly quickly because only the deltas are exchanged. Larger devices, such as local or cloud servers, connected to the mesh through the internet, should be responsible for synchronizing as much data as possible, ensuring data access and visibility for users who are not directly connected to the local mesh.
When considering the latency aspect of synchronization, a peer-to-peer mesh network must make it easy for other devices to join and leave when they need to. Within this model, it is key to ensure that all devices have input from the same data source. However, this poses a great mathematical challenge as the mesh network topology changes over time. So, it is important for these mesh networks to be flexible without needing to have the complete full history of a database table to write or read the latest value. Therefore, creating an ad-hoc network.
For this to be successful, the peer-to-peer mesh network must understand that devices update frequently but at different times, so it must take into account the incoming data with different schemas. In this way, even if a device is offline and, therefore, outdated, it should still be able to read new data and sync. The way to do this lies in how the network works with a device in a reliable order of changes that can be inspected, which also includes incorporating metadata about schema changes over time.
Developers need the proper tools to be able to purposely build cloud-optional apps. When there is an interruption, all devices must see the same query results given the same set of changes, even if the changes arrive in a different order. It is challenging for development teams to create a dependable, offline-first, peer-to-peer datastore that syncs data in a partially connected mesh. However, as cloud-optional capabilities become more attractive, there will be growing demand for a complete end-to-end solution that combines the best of cloud software with the best of peer-to-peer software.
This article was produced as part of TechRadarPro’s Expert Insights channel where we feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/news/submit-your-story-to-techradar-pro
Johnblack Kabukye struggles to explain to his colleagues back home in Uganda why he’s doing a two-year stint as a postdoctoral researcher in Sweden. “If you say you’re doing a master’s or a PhD, it’s clear what that means,” says the digital-health specialist, who worked as a physician for a decade before switching to research. But a postdoc? “It’s not a thing that is understood,” he says.
The skills he’s learning at Stockholm University while building electronic health tools tailored to patients’ needs are certainly useful for his job as a physician and informatician at the Uganda Cancer Institute (UCI) in Kampala. But the postdoc format itself — a short-term position designed to bridge the gap between doctoral student and tenured academic — makes little sense in Uganda, where it is common to have a permanent teaching job at a university before even embarking on a PhD.
“I have not heard of a single postdoc opportunity in Uganda,” Kabukye says.
Career resources for African scientists
That could soon change across Africa. The number of people gaining PhDs in the continent is growing, and so is the need for postdoctoral employment. “There is definitely greater awareness of the postdoc position, and more and more postdocs,” says Gordon Awandare, pro-vice-chancellor in charge of academic and student affairs at the University of Ghana in Accra.
But as the continent’s postdoctoral employment needs have grown, so too have fears that the problems created by a proliferation of postdoc positions in other parts of the world — which critics say trap young researchers in a cycle of poorly paid, short-term positions with no job security — could also arise in African countries.
Breaking ground
Postdoc frustration is a recurring theme in studies that look at early-career researchers. Two global surveys of postdoctoral students by Nature, one published in 2020 and the other last year, found that more than one-third of respondents were dissatisfied with their lot. A lack of job security, career-advancement opportunities and funding were the most-cited reasons.
Nature’s surveys underscore the dearth of postdoctoral researchers in Africa. Of the 3,838 postdocs surveyed in June last year, only 91 were based on the continent. The number of respondents (who were self-selecting) were too few, and too geographically concentrated in three countries — South Africa, Nigeria and Egypt — to be viewed as representative of the continent. Yet, they offer tantalizing glimpses of an emerging segment of the global research workforce.
For example, African postdocs were older than the global average, with more than 40% aged 41 or older. They were also more likely to be doing their postdoc in their home country (68% in Africa compared with 39% globally) and they were much less likely, than the global average, to be employed on fixed-term fellowships or contracts (see ‘Employment matters’). Their pay also stands out: 60% said they earned less than US$15,000 per year — the lowest option survey-takers could tick, and a fraction of what most postdocs are paid in Europe and North America. Lower costs of living play some part in the lower salaries, but not enough to justify the gap (see ‘A continental shift’).
Postdocs in Africa were also more likely to report having a second job alongside their postdoc than were other respondents, on average (33% of respondents in Africa, compared with 10% of respondents overall). The most common reason was to provide extra income (71%), while 57% said their second job gave their skills and career prospects a boost. However, notes Awandare, the tendency of many African postdocs to have permanent academic positions before becoming a postdoc could be a confounding factor in this measure.
Yet, and perhaps surprising given their low pay, Africa-based postdocs were the most optimistic about their futures of all respondents from the geographical regions represented. Overall, 64% of Africa-based respondents reported that they felt positive about their future job prospects, compared with 41% globally. Postdocs in Africa were twice as likely to say that their postdoc roles were better than they imagined (25% compared with 12% overall). And 42% of respondents in Africa felt that they had better prospects than previous generations of postdocs, far exceeding the 15% global average.
That optimism makes sense to Awandare, who thinks that postdocs in his country might feel more important than do their peers who work in large laboratories overseas. In addition to his leadership role at the University of Ghana, he founded and runs the West African Centre for Cell Biology of Infectious Pathogens at the university. He says postdocs at the centre are treated the same as faculty members. “In some advanced institutions, they wouldn’t get that recognition and status,” he notes.
And even though their salaries are low by international standards, postdocs at his centre can be better paid than entry-level permanent university staff who only teach, he says. This is because postdocs tend to be paid out of lucrative international grants. “Ten to fifteen years ago, many of these positions would have been overseas — but now funders, to their credit, increasingly provide positions on the continent,” he says.
A different set-up
Employment structures also differed between Africa and the rest of the world, according to Nature’s survey. Although similar proportions of postdocs were employed in academia in Africa as they were globally (around 90%), the proportion of part-time postdocs was higher in Africa — 12% compared with the global average of 5%. One of them is Felista Mwingira, a parasitologist at the University of Dar es Salaam in Tanzania. She exemplifies how African early-career researchers have been forging ahead in their research careers in the absence of a formal structure of postdoc positions.
Mwingira obtained her PhD in 2014 from the University of Basel in Switzerland at the age of 33 — which she says is very young for researchers in Tanzania. By the time she started her studies, she was already permanently employed by her university in Tanzania, and was able to return to that post after finishing her PhD. Back home, she could take three months paid maternity leave for each of her two children, born four years apart. And although juggling pregnancies and bringing up children with the demands of an academic career was a challenge, it meant she had job security — something postdocs at the same stage in their lives in other parts of the world often lack.
Falling behind: postdocs in their thirties tire of putting life on hold
Mwingira’s work after her PhD was not technically a postdoc. But as her children got older, she sought out a mentorship arrangement at her university that provides her with research training and, sometimes, extra money from the projects she works on. It’s not a formal postdoc, but she hopes it will help her to attain the publication ‘points’ required in the Tanzanian university system to progress up the academic career ladder — something that does not depend on more-senior positions becoming available. She hopes to be promoted in the near future, but says she would also like to embark on a full-time postdoc position to “sharpen my scientific skills”.
So far, Mwingira considers herself lucky. Her children are now four and eight, and while she says that her life as an early-career academic still has ups and downs, she is thankful for the stability she has enjoyed so far in her career. “I think that I’m better off compared to postdocs in high-income countries.”
That feeling of being better off than people elsewhere certainly does not translate to sub-Saharan Africa’s most prominent research nation: South Africa. There, postdoc numbers have been rising for a couple of decades, growing from around 300 in 1999 to nearly 3,000 in 2019 (ref. 1), and national surveys reveal postdoc frustrations that mirror those raised globally, with some country-specific gripes to boot.
Heidi Prozesky is a research scholar at the Centre for Research on Evaluation, Science and Technology at Stellenbosch University. She is one of the people behind South Africa’s first PhD tracer study, published in its final form in July 2023, which tracked the whereabouts of nearly 6,500 PhDs who had graduated in the country between 2010 and 2019. That survey found that around 20% had accepted at least one postdoctoral fellowship, either at home or abroad, on completing their PhDs, with a steady growth seen over the two decades. The postdocs spent a median of three years in the position, although one-quarter reported spending more than four years. One-third reported having accepted more than one postdoc — often, they said, because other work was not available.
Career resources for postdoctoral researchers
A common refrain in the South African survey, which echoes the findings of Nature’s global surveys, is that postdocs feel like they are in limbo: neither students nor staff. In reality, postdocs in South Africa are technically students. This saves them from paying tax on their income, which are stipends, not salaries. But this designation also breeds resentment, because it means postdocs are treated like students: they can’t apply for grants and typically have no funding to travel to conferences or attend workshops.
In addition to the lack of opportunities, postdoc pay in South Africa is low compared with living costs. Last year, the National Research Foundation’s non-taxable postdoc stipends started at 200,000 rand (US$10,700). Female postdocs are allowed up to four months paid maternity leave. However, basic private medical insurance does not come as standard, meaning that postdocs have to pay for it out of their stipends if they want to avoid state health care, which many people in South Africa view as woefully inadequate. The stories of some postdocs “would make you cry”, says Palesa Mothapo, who heads research support and management at Nelson Mandela University in Port Elizabeth, South Africa. “These people have PhDs. And they end up going hungry.”
Growing pains
South Africa’s predicament stems partly from bottlenecks in the academic careers system. The number of people with a PhD graduating annually more than tripled between 2000 and 2018, increasing the demand for postdoctoral work. Postdoc positions have also increased, but further up the career ladder, the number of roles has been static. A study published this year1 in the South African Journal of Science found that the number of postdoc positions grew ten times faster between 2007 and 2019 in the country than did the growth in entry-level permanent jobs in academia.
Palesa Mothapo at Nelson Mandela University in Port Elizabeth, South Africa, says there needs to be more discussion around transferable skills for African postdocs.Credit: Stefan Els
But many also view South Africa’s postdoc malaise as a consequence of incentive structures in the country that place a premium on research publications. Postdocs have become cheap, low-commitment hires for universities that want to boost their output of research publications, which in South Africa earn the host institutions or departments cash subsidies from the government. Postdocs often have publication targets written into their appointments, Mothapo says. “But those papers don’t translate to money for the postdoc. It goes to the institution, to the host.”
There is some cause for cheer. Last December, the National Research Foundation announced it would raise its minimum annual postdoc stipend to 320,000 rand per year for new fellowships from 2024. But simply increasing postdoc stipends is unlikely to create more academic positions for postdocs who are looking for more job security. And the bottleneck seems to be worse for some groups. According to Prozesky, South Africa attracts a lot of postdocs from the rest of the African continent. Most come with the expectation that it will lead to a permanent job. The PhD tracer study found that many people from the rest of Africa end up disillusioned and feeling discriminated against. They struggle to move on from the postdoc status, and can face long delays in visa approvals when moving between posts. “They call it academic xenophobia,” says Prozesky.
Charles Teta, a Zimbabwean environmental chemist who did two postdocs in South Africa after a PhD in his home country, says that he noticed that South African citizens were less likely to take the postdoc route than were immigrants like him. “South Africans are more likely to get lectureship posts,” without having any postdoc experience, he says. In addition, a growing number of funding streams are not open to non-citizens — even those who are permanent residents. Eventually, those restrictions cause people to leave, he says.
Teta left South Africa last year to cover the maternity leave of an environmental-science lecturer at Queen Mary University of London. There, he enjoys the opportunity to teach — something he wasn’t expected to do during his postdocs. It’s been a happy choice so far, and he hopes to find another, similar position when his current one ends. He doesn’t miss the research treadmill, which, he says, “did not translate to mental and financial well-being”.
A call for creativity
Mothapo says that the rigid focus on research in South African postdoc roles is part of their problem. “The universities are not creative,” she says. Because postdocs are limited in how they can teach, and can’t apply for their own funding, she notes, they are missing out on learning skills that are beneficial for staying in academia, and that could open up alternative career paths in industry.
More-creative programmes have been trialled across the continent. Since 2019, the US National Institutes of Health (NIH), the Bill & Melinda Gates Foundation in Seattle, Washington and the African Academy of Sciences have been running the African Postdoctoral Training Initiative (APTI). The programme combines a two-year postdoc at a NIH institute in the United States with a two-year research grant that fellows can take back home to build their own research programmes. Notably, it is open only to researchers who have permanent positions already.
Postdoc career optimism rebounds after COVID in global Nature survey
Daniel Amoako-Sakyi, an immunologist at the University of Cape Coast, Ghana, embarked on an APTI fellowship in late 2023. He is a postdoc in mid-life, and the fellowship has proved to be a good fit. He is a few months into his position at the National Institute of Allergy and Infectious Diseases in Bethesda, Maryland, where he will spend the next two years looking at biological reasons for the variance in efficacy seen in new malaria vaccines. His 15-year-old daughter has enrolled in a US high school, and his spouse, a fellow academic, aims to split her time between the United States and Nigeria.
In Bethesda, Amoako-Sakyi has none of the resource constraints that limit him in Ghana. Antibodies that would take months to ship to his home country arrive on his doorstep overnight. He expects the opportunity will supercharge his career, and hopes he’ll be able to take on some postdocs of his own when he returns home. He doesn’t expect it will be difficult to find them. “I think most researchers are looking for the right environment to flourish,” he says.
What comes next?
There are few certainties about the future of African postdocs. Those who spoke to Nature hope that their postdoc training will accelerate their careers — by helping them to win grants, get promotions and expand their research networks. In Uganda, Kabukye hopes to have organized funding and collaborators by the end of his postdoc so that he can carry on his research designing and implementing digital-health tools in resource-constrained settings. “Ideally, I would have positions at the UCI and at another university, to foster collaboration and exchange,” he says.
Physician Johnblack Kabukye from Uganda is doing a postdoc building electronic health tools at Stockholm University in Sweden.Credit: Johnblack Kabukye
However, with most of the continent’s research funding still coming from sources outside Africa — with the exception of a handful of countries, such as South Africa and Egypt — it’s likely that foreign funding will keep driving the creation of postdoc opportunities. And that can mean the positions aren’t always tailored to local needs.
Mothapo says that she often hears research funders talk about the need to create more postdoc positions. However, there is not enough discussion around the particular needs that African postdocs will have, especially the transferable skills that they will need if they want to transition to sectors such as industry. “I’m worried about their destinations,” she says.
Mwingira echoes her concern. She thinks that more formalized postdocs in Tanzania could lead to bottlenecks in the training system, as has been seen in South Africa and elsewhere. “Those problems will arise in Tanzania, too, but worse, because of the low salaries,” she says.
But Amoako-Sakyi does not think that the creation of more African postdocs has to result in frustration as they compete for rare academic posts. Many might already be employed by universities at that point in their careers. A postdoc could allow them to win grants from funders so that they can set up their own research groups and create opportunities for the next generation. He also thinks that the biotechnology industry in countries such as Ghana will grow, further increasing the demand for researchers in the country.
Nor does Amoako-Sakyi think that African postdocs need to end up in the same negative landscape that postdocs occupy elsewhere in the world. Such fears are not unfounded, he says, because concepts are often brought to the continent and adopted without thinking about the local context. But as his own fellowship shows, there are ways to tailor postdocs to African settings. “We should be very intentional about how we do it and try to correct old mistakes.”
We have covered plenty of projects that have been created over the past few months using the new Microsoft AutoGen framework which was quietly rolled out to GitHub. offering a framework that enables the development of LLM applications using multiple agents, capable of communicating with each other to solve tasks. The beauty of AutoGen agents is that they are customizable, conversable, and seamlessly allow human participation. They can operate in various modes that employ combinations of LLMs, human inputs, and tools.
If you’ve ever been captivated by the idea of automating complex workflows using artificial intelligence, you will be pleased to know that AutoGen is at the forefront of this emerging landscape. Imagine a world where your projects are not just assisted by a single language model, but an entire team of specialized AI agents, conversing amongst themselves and executing tasks at an unprecedented scale. Intrigued? Let’s delve deeper into how you can build a virtual workforce of AI helpers using AutoGen and GPT-4.
“GPT-4, the latest milestone in OpenAI’s effort in scaling up deep learning. GPT-4 is a large multimodal model (accepting image and text inputs, emitting text outputs) that, while less capable than humans in many real-world scenarios, exhibits human-level performance on various professional and academic benchmarks.”
Team of AI agents working together
At the core of AutoGen lies its capability to simplify the orchestration, automation, and optimization of intricate workflows involving language models like GPT-4. While there are other contenders in this space—think MetaGPT or ChatDev—AutoGen stands out for its focus on multi-agent conversations. What this means is that you can have several agents, each programmed for specific roles or tasks, working in concert. Not only does this make the system more robust by offsetting individual limitations of single agents, but it also enables a level of customization that is hard to match.
Other articles we have written that you may find of interest on the subject of Microsoft’s AutoGen AI Agent framework :
Microsoft AutoGen AI agent framework
If you are wondering how to adapt this to suit your specific needs, AutoGen provides tools for customizing the conversational patterns of your agents. Whether you’re considering one-to-one, multi-agent, or even complex tree-like conversational topologies, it’s all within reach. You get to decide the number of agents involved and the degree to which they can converse autonomously. This is highly beneficial for applications requiring a diversity of conversational styles and structures, from customer service to project management and beyond.
AutoGen is versatile in its application, able to accommodate a multitude of use-cases across various sectors. Be it healthcare, finance, or retail, the framework has pre-built, working systems that can be adapted to different complexities and requirements. This is an invaluable asset for those wanting to integrate AI into specialized domains without reinventing the wheel.
In terms of technical infrastructure, AutoGen brings several advantages to the table. It offers enhanced performance tuning options, API unification, and caching functionalities. Advanced features like error handling, multi-config inference, and context programming are also part of the package. Essentially, you get a plethora of utilities to ensure that your virtual workforce performs optimally.
How to build a virtual AI workforce
If you’re eager to dive in, the easiest entry point is through Github Codespace. Simply copy the sample OAI_CONFIG_LIST to the /notebook folder, rename it to OAI_CONFIG_LIST, and set the configurations as needed. From there, you’re all set to explore and experiment with the example notebooks. Full instructions on how to use Microsoft’s AutoGen and Codespaces can be found over on GitHub.
“Create a codespace to start developing in a secure, configurable, and dedicated development environment that works how and where you want it to.”
While automating tasks is compelling, there are instances when human intuition and expertise cannot be replicated by machines. Recognizing this, AutoGen is designed to seamlessly integrate human input and feedback into the system. You, or any other human user, can interact with the agents, guiding them towards better solutions or intervening when necessary.
So there you have it—an intricate yet user-friendly guide to creating a virtual team of AI helpers, effortlessly amalgamating the individual strengths of multiple agents into a coherent and efficient workforce. If you are invested in leveraging AI for complex problem-solving, AutoGen, coupled with GPT-4, offers a promising avenue to make this a reality.
Filed Under: Guides, Top News
Latest timeswonderful Deals
Disclosure: Some of our articles include affiliate links. If you buy something through one of these links, timeswonderful may earn an affiliate commission. Learn about our Disclosure Policy.
Building your very own AI workforce of virtual helpers or AI agents is a lot easier than you might think. If you have a computer running over 8 GB of RAM you can easily install your own personal AI using Ollama in just a few minutes. Once installed Ollama allows you to easily install a wide variety of different AI models however you will need more RAM to run the larger models such as Llama 2 13B. As large language models tend to consume a significant amount of RAM. Although if you would like to get more advanced and improve the performance of your LLM this can be done using StreamingLLM.
Microsoft’s AutoGen has emerged as a powerful tool for creating and managing large language model (LLM) applications. This innovative framework enables the development of LLM applications using multiple agents that can converse with each other to solve tasks. The agents are customizable, conversable, and seamlessly allow human participation. They can operate in various modes that employ combinations of LLMs, human inputs, and tools.
AutoGen was developed by Microsoft with the aim of simplifying the orchestration, automation, and optimization of complex LLM workflows. It maximizes the performance of LLM models and overcomes their weaknesses. This is achieved by enabling the building of next-gen LLM applications based on multi-agent conversations with minimal effort.
Build a team of AI assistants using AutoGen
Watch the video below to learn more about building your very own AI workforce to help you power through those more mundane tasks allowing you to concentrate on more important areas of your life or business. Follow the step-by-step guide kindly created by the team over at WorldofAI.
Previous articles you may find of interest on Microsoft’s AuotGen framework :
One of the key features of AutoGen is its ability to create multiple AI agents for collaborative work. These agents can communicate with each other to solve tasks, allowing for more complex and sophisticated applications than would be possible with a single LLM. This multi-agent conversation capability supports diverse conversation patterns for complex workflows. Developers can use AutoGen to build a wide range of conversation patterns concerning conversation autonomy, the number of agents, and agent conversation topology.
AutoGen’s architecture is highly customizable and adaptable. Developers can customize AutoGen agents to meet the specific needs of an application. This includes the ability to choose the LLMs to use, the types of human input to allow, and the tools to employ. Furthermore, AutoGen seamlessly allows human participation, meaning that humans can provide input and feedback to the agents as needed.
AutoGen features
Multi-Agent Conversations: Enables development of LLM applications using multiple, conversable agents that interact to solve tasks.
Customizable and Conversable Agents: Agents can be tailored to fit specific needs and can engage in diverse conversation patterns.
Human Participation: Seamlessly integrates human inputs and feedback into the agent conversations.
Versatile Operation Modes: Supports combinations of LLMs, human inputs, and tools for varied use-cases.
Performance and optimization
Workflow Simplification: Eases the orchestration, automation, and optimization of complex LLM workflows.
Performance Maximization: Utilizes features to overcome LLM weaknesses and maximize their performance.
API Enhancement: Provides a drop-in replacement for openai.Completion and openai.ChatCompletion with additional functionalities like performance tuning and error handling.
Application scope
Diverse Conversation Patterns: Supports a variety of conversation autonomies, number of agents, and topologies.
Wide Range of Applications: Suits various domains and complexities, exemplified by a collection of working systems.
Technical details
Python Requirement: Needs Python version >= 3.8 for operation.
Utility Maximization: Optimizes the use of expensive LLMs like ChatGPT and GPT-4 by adding functionalities such as tuning, caching, and templating.
Installation of AutoGen requires Python version 3.8 or higher. Once installed, AutoGen provides a collection of working systems with different complexities. These systems span a wide range of applications from various domains and complexities, demonstrating how AutoGen can easily support diverse conversation patterns.
AutoGen also enhances the capabilities of existing LLMs. It offers a drop-in replacement of openai.Completion or openai.ChatCompletion, adding powerful functionalities like tuning, caching, error handling, and templating. For example, developers can optimize generations by LLM with their own tuning data, success metrics, and budgets. This feature helps maximize the utility out of the expensive LLMs such as ChatGPT and GPT-4.
In terms of its potential, AutoGen stands out in comparison to other AI agents. Its ability to support diverse conversation patterns, its customizable and conversable agents, and its seamless integration of human participation make it a powerful tool for developing complex LLM applications.
Microsoft’s AutoGen is a groundbreaking tool that enables the creation and management of large language model applications. Its multi-agent conversation framework, customizable and conversable agents, and seamless integration of human participation make it a powerful tool for developers. Whether you’re looking to optimize the performance of existing LLMs or create complex, multi-agent applications, AutoGen offers a robust and flexible solution.
AutoGen is an open-source, community-driven project under active development (as a spinoff from FLAML, a fast library for automated machine learning and tuning), which encourages contributions from individuals of all backgrounds.
Filed Under: Guides, Top News
Latest timeswonderful Deals
Disclosure: Some of our articles include affiliate links. If you buy something through one of these links, timeswonderful may earn an affiliate commission. Learn about our Disclosure Policy.