Mediaboss Marketing

ChatGPT has sparked speculation about artificial general intelligence. But the next real phase of AI will be in specific domains and contexts.

The relentless hype surrounding generative AI in the past few months has been accompanied by equally loud anguish over the supposed perils just look at the open letter calling for a pause in AI experiments. This tumult risks blinding us to more immediate risks think sustainability and bias and clouds our ability to appreciate the real value of these systems: not as generalist chatbots, but instead as a class of tools that can be applied to niche domains and offer novel ways of finding and exploring highly specific information.

This shouldnt come as a surprise. The news that a dozen companies have developed ChatGPT plugins is a clear demonstration of the likely direction of travel. A generalized chatbot wont do everything for you, but if youre, say, Expedia, being able to offer customers a simple way to organize their travel plans is undeniably going to give you an edge in a marketplace where information discovery is so important.

Whether or not this really amounts to an iPhone moment or a serious threat to Google search isnt obvious at present while it will likely push a change in user behaviors and expectations, the first shift will be organizations pushing to bring tools trained on large language models (LLMs) to learn from their own data and services.

Dont settle for half the story.Get paywall-free access to technology news for the here and now.

And this, ultimately, is the key the significance and value of generative AI today is not really a question of societal or industry-wide transformation. Its instead a question of how this technology can open up new ways of interacting with large and unwieldy amounts of data and information.

OpenAI is clearly attuned to this fact and senses a commercial opportunity: although the list of organizations taking part in the ChatGPT plugin initiative is small, OpenAI has opened up a waiting list where companies can sign up to gain access to the plugins. In the months to come, we will no doubt see many new products and interfaces backed by OpenAIs generative AI systems.

While its easy to fall into the trap of seeing OpenAI as the sole gatekeeper of this technology and ChatGPT as the go-to generative AI tool this fortunately is far from the case. You dont need to sign up on a waiting list or have vast amounts of cash available to hand over to Sam Altman; instead, its possible to self-host LLMs.

This is something were starting to see at Thoughtworks. In the latest volume of the Technology Radar our opinionated guide to the techniques, platforms, languages and tools being used across the industry today weve identified a number of interrelated tools and practices that indicate the future of generative AI is niche and specialized, contrary to what much mainstream conversation would have you believe.

Unfortunately, we dont think this is something many business and technology leaders have yet recognized. The industrys focus has been set on OpenAI, which means the emerging ecosystem of tools beyond it exemplified by projects like GPT-J and GPT Neo and the more DIY approach they can facilitate have so far been somewhat neglected. This is a shame because these options offer many benefits. For example, a self-hosted LLM sidesteps the very real privacy issues that can come from connecting data with an OpenAI product. In other words, if you want to deploy an LLM to your own enterprise data, you can do precisely that yourself; it doesnt need to go elsewhere. Given both industry and public concerns with privacy and data management, being cautious rather than being seduced by the marketing efforts of big tech is eminently sensible.

A related trend weve seen is domain-specific language models. Although these are also only just beginning to emerge, fine-tuning publicly available, general-purpose LLMs on your own data could form a foundation for developing incredibly useful information retrieval tools. These could be used, for example, on product information, content, or internal documentation. In the months to come, we think youll see more examples of these being used to do things like helping customer support staff and enabling content creators to experiment more freely and productively.

If generative AI does become more domain-specific, the question of what this actually means for humans remains. However, Id suggest that this view of the medium-term future of AI is a lot less threatening and frightening than many of todays doom-mongering visions. By better bridging the gap between generative AI and more specific and niche datasets, over time people should build a subtly different relationship with the technology. It will lose its mystique as something that ostensibly knows everything, and it will instead become embedded in our context.

Indeed, this isnt that novel. GitHub Copilot is a great example of AI being used by software developers in very specific contexts to solve problems. Despite its being billed as your AI pair programmer, we would not call what it does pairing its much better described as a supercharged, context-sensitive Stack Overflow.

As an example, one of my colleagues uses Copilot not to do work but as a means of support as he explores a new programming language it helps him to understand the syntax or structure of a language in a way that makes sense in the context of his existing knowledge and experience.

We will know that generative AI is succeeding when we stop noticing it and the pronouncements about what it might do die down. In fact, we should be willing to accept that its success might actually look quite prosaic. This shouldnt matter, of course; once weve realized it doesnt know everything and never will that will be when it starts to become really useful.https://wp.technologyreview.com/wp-content/uploads/2022/04/Thoughtworks_Video_ContributedArticle_April2022.mp4Provided by Thoughtworks

This content was produced by Thoughtworks. It was not written by MIT Technology Reviews editorial staff.

Follow this link:

The future of generative AI is niche, not generalized - MIT Technology Review

I recently attended the 2023 annual meeting of the American Roentgen Ray Society (ARRS), one of the major professional societies for radiologists and medical imaging specialists. As expected, one of the hot topics was artificial intelligence (AI) and expected impact on radiologists in particular, as well as medical practitioners in general.

Although I could not attend all of the numerous lectures, panel discussions, and research presentations on AI, I did learn of many exciting developments as well as areas of both opportunity and concern. In this column, Id like to share some thoughts on how AI will affect patients and physicians alike in the short-to-medium term future.

(Note: This discussion will be confined to so-called narrow AI to accomplish particular medical tasks, rather than artificial general intelligence or AGI that can simulate general human cognition. Ill leave the debate over whether a sufficiently advanced AI will exterminate humanity to others.)

1) AI will play an increasingly greater role in medical care, in ways both obvious and non-obvious to patients.

In my own field of radiology, AI will be used to enhance (but not yet replace) human radiologists making diagnoses from medical images. There are already FDA-approved AI algorithms to detect subtle internal bleeding within the brain or potentially fatal blood clots (pulmonary embolism) within the arteries of the lung.

When properly used, these algorithms could alert the human radiologists that a patients scan has one of these life-threatening abnormalities and bump the case to the top of the priority queue. This could significantly shorten the time between the scan and the appropriate treatment and thus save lives. (See this paper by Dr. Kiran Batra and colleagues from University of Texas Southwestern Medical Center for one example of the time savings achieved by AI.)

AI can also be used to enhance medical care in ways not directly related to rendering diagnoses. For instance, developers are working on physician co-pilot software that can sift through a patients medical records and extract the information most relevant for the patients upcoming visit to the radiology department (or internal medicine clinic, etc.). This could save the practitioners valuable time during each patient visit.

Robotic physician holding stethoscope

getty

2) The AIs are still not perfect, and human physicians will still need to have the final say in diagnoses and treatments.

For example, AIs are pretty good in detecting early breast cancer in mammogram images, but still make errors. (Often they make errors humans dont, and vice versa.) This makes AI great as an assistant to the human radiologist, but not (yet) a viable replacement.

Thus, we will see an interesting period of time where human physician-plus-AI will perform better than either human alone or AI alone. At some point in time, I predict that AI-assisted medicine will become standard of care and physicians who do not incorporate AI into their daily practices could open themselves to lawsuits for practicing substandard care.

3) As AIs get better, humans may start to over-rely on them.

This phenomenon is known as de-skilling. As an analogy (made by Dr. Charles Kahn of University of Pennsylvania in one of the ARRS panel discussions), suppose we develop self-driving automobiles that could handle most traffic conditions, but still required a human driver to take the wheel in emergencies. As AIs got increasingly better and the need for human intervention became less frequent, we human drivers could easily become complacent and lose good driving-related cognitive habits and reflexes.

If a partially-automated car going 70 mph on the highway suddenly alerted a human driver who hadnt truly driven in the past year to take over because of icy conditions ahead, things could go badly.

Similarly, if a human radiologist lets their cancer detection skills go rusty, they could run into trouble when the medical images included complex visual features beyond the ability of the AI to accurately parse.

My own personal approach will be to think of the AI as a tireless-but-quirky medical student constantly asking questions like, Could that squiggle be a cancer? How about this dark line is it a fracture? Could this dot be a small blood clot? An inquisitive human medical student can keep experienced doctors on their toes in a good way, and the same could be true for an AI.

4) AI could take over some interactions with patients that currently require human medical personnel.

Were probably not too far from reaching the point that a LLM (Large Language Model) AI like ChatGPT could take a radiology report written in medical jargon and translate it into terms understandable to non-physicians and possibly even answer follow-up questions about the significance of the findings.

A recent article by Ayers and colleagues in JAMA Intern Med compared how AI chatbots and human physicians responded to patient medical questions offered on social media. According to the judges (who were blinded as to the author of the answers), the chatbot answers were considered better both in terms of information quality and empathy than the human physicians answers!

The use of artificial intelligence in medicine is a rapidly evolving field, and Ive only scratched the surface of the exciting work being done. Given the rapid pace of developments, I dont know what things will look like in 5 months, let alone in 5 years. But Im glad to be alive during this time of potentially massive innovation (and admittedly potentially uncomfortable upheaval). For now, I remain optimistic that AI could be an enormous boon for patients and physicians alike.

I am a physician with long-standing interests in health policy, medical ethics and free-market economics. I am the co-founder of Freedom and Individual Rights in Medicine (FIRM). I graduated from University of Michigan Medical School and completed my residency in diagnostic radiology at the Washington University School of Medicine in St. Louis (where I was also a faculty member). I'm now in private practice in the Denver area. All my opinions are my own, and not necessarily shared by my employer.

Read this article:

Promises, Perils, And Predictions For Artificial Intelligence In Medicine: A Radiologists Perspective - Forbes

In less than six months, ChatGPT has become a household name. The AI service can write paragraphs, essays, speeches and fill in exams. So many people have flocked to the chatbot for a glimpse of its power that the servers have to be shut down at times. It is a tour de force of artificial intelligence.

ChatGPT was developed by OpenAI, an artificial intelligence company founded in 2015 with a mission to ensure that artificial general intelligence benefits all of society, according to OpenAIs website.

The human-like chatbot can answer almost any question the user provides, and it has been trained to respond as a human would.

When asked about its pros and cons, ChatGPT said, Its important to note that while I can be a helpful tool for certain tasks, human judgment and critical thinking should always be exercised when interpreting and using the information generated by AI systems like me.

With ChatGPTs capabilities, it comes as no surprise students have been tempted to consult it for assistance with their assignments, especially in English courses. However, some Dallas College faculty warn students of ChatGPTs downsides.

A lot of people have heard of it but arent sure exactly what it does, or what it doesnt do, Marylynn Patton, Dallas College El Centro Campus ESOL curriculum chair, said. It doesnt do everything.

Patton recently presented on ChatGPT at a national Teaching English to Speakers of Other Languages conference where she spoke about ways ChatGPT can be used as an educator, and how to detect whether something has been written using ChatGPT.

She said since its release in November 2022, ChatGPT has greatly improved. Where it used to score mid-range on AP exams and Bar exams, it is now scoring higher than 90%, Patton said.

One area where the AI falters is in English and literature courses. Patton said ChatGPT is scoring a two on the AP English exam, which is below college level. [ChatGPT] is not highly qualified, Patton said.

In the lower-level skills like read, respond, summarize, [with] those things it can do pretty well, Patton said. Its the higher level, the critical thinking, the evaluating materials, giving reactions to things. Those are the things that it cannot do.

Dallas College does not have an official stance on ChatGPT yet, Patton said. But she urges instructors to sway students from resorting to ChatGPT. She suggests that teachers discuss the AI chatbot with students. Talk about the ethical side of it, how it could be used or ways that its being abused, Patton said.

On the upside, ChatGPT is useful for providing formats for essays and letters, Patton said.

English professor Kendra Unruh said she is changing how she formats assignments for her students. One thing she does is have students write a rough draft in class so she has something on which to base her students writing. If a student turns in the final draft and their writing style veers from the rough draft, it will be obvious they did not write the essay.

Unruh has also updated her discussion board posts to directly ask students what they thought about a topic. She said she tries to make the process personal enough that an AI can not reproduce the results.

Patton related ChatGPT to a calculator. It will be the new calculator for writing, Patton said. In math, you learn your basics and then after you learn your basics you can use the calculator.

Read more from the original source:

AI seeps into coursework The Brookhaven Courier - Brookhaven Courier

Sitting on the front lawn of my cottage in the forest sanctuary of Binsar in the Kumaon hills, I am struggling to pick my way between the arguments on whether the further development of Artificial General Intelligence should be paused or not.

Regular readers will know I write an occasional column from the remote fastness of this sanctuary. My cottage can only be approached by foot. The nearest market is a 45-minute drive and I have to haul up the provisions required for the duration of my stay; there is no grid electricity (I have installed solar) and no running water (I source rainwater from tanks). But there is connectivity. I have access to 4G telephony and WiFi. I am therefore able to keep track of worldly affairs.

Were I in Mumbai or Delhi, the debate triggered by the release of the neural language model ChatGPT 4 would have engaged my intellect. But I am not in those cities. I am instead perched in splendid isolation I forgot to add my nearest neighbour is a 20-minute walk on a promontory overlooking dense oak forests with the Himalayan peaks, Nanda Devi and Trishul in the foreground, the twitter of birds as the only ambient background sound and the words of the Nobel Prize prize-winning Russian novelist Aleksandr Solzhenitsyn, man is but an insignificant creature of creation reverberating in my brain. It is not my mind driving my thoughts. It is my senses. And that is why I am not sure where to pitch my flag.

I have read excerpts of the letter coordinated by the Future of Life Institute and signed by apparently thousands of scientists, technocrats, businessmen, academics and others (the exact number of signatories is not known as there are many forged signatures ) calling for a six-month pause in the further development of neural language models. The signatories include Elon Musk who, ironically, was a co-founder of Open AI, the inventor of ChatGPT, but who sold his shares after a tiff with the other founders Steve Wozniak, the cofounder of Apple and the Israeli Philosopher and author of Homo Deus, Yuval Noah Harari. The central message of the letter is that further unconstrained development of such language models could create human competitive intelligence that if not circumscribed by governance protocols could pose a profound risk to humanity. Further work should therefore be halted until such protocols are in place.

This letter reminded me of the comment made by Robert Oppenheimer, the Director of the Manhattan Project that designed the atomic bomb when he became aware of the destructive potential of his creation. Now, I am become Death, the destroyer of worlds. (a loose translation of a verse from the Bhagwad Gita). Oppenheimer spent much of the rest of his professional life lobbying to contain the fallout.

I have also read the counters to this letter. Many have dismissed it as paranoiac hype. Some have argued that this is not an unexpected reaction. Every technological transformation has triggered opposition by vested interests. The abbott Johannes Trithemius opposed the invention of the printing press by Johannes Gutenberg in 1436 because he thought it would make monks lazy. The word of God needs to be interpreted by priests not spread about like Dung, he said. The industrial Luddites of the early 19th century protested the mechanisation of the knitting loom out of concern for the livelihood of skilled weavers. More contemporaneous, many have forewarned against the adverse impact on jobs, data privacy and individual rights of the digital revolution. There is much in these reactions, but had they led to the stoppage of further technological developments, society would have been worse off. On that, there is no doubt. Some have also adduced the geopolitical argument that a pause will grant China an open sesame on AI and that would be a setback for the rules-based, liberal, international order.

I reflect on these arguments but my thoughts are not clear.

At one level, I am drawn to the implicit message in the letter that enough is enough. That whilst human ingenuity has indeed improved the nature of our daily lives, it has also brought us to the brink of a planetary catastrophe. There is no doubt that one reason I am able to look across verdant hills is that as a declared sanctuary, Binsar, has been protected from the ravages of industrialisation. I am also concerned (without being able to put a finger on the precise reasons for this concern) that if the motive force driving the phenomenally talented is personal profits rather than public welfare and if there are no protocols or guard rails, then through the self-reinforcing momentum of creativity, a situation may well arise wherein the creator loses control over his creation. Decision making would then pass onto the levers of inanimate, albeit intelligent machines. The ramifications could be frightening.

At another level, however, I wonder how in the absence of technological progress we can get back on the rails of sustainable development. One reason the world is still hopeful of tackling global warming is that technology has rendered clean energy a competitive alternative to fossil fuels. Further technological progress should enable the sequestration of carbon from the atmosphere. That would be a transformational step in the journey towards decarbonisation. I also wonder about the practicality of getting individuals to pause their innate instinct to experiment, innovate and create. Would that not require upending the liberal values that place individual rights at the centre of public governance?

As I said, I am not clear where to pitch my flag. But I sense the real problem is not the unbridled momentum of AI. It is the international communitys inability to look beyond narrow jingoistic interests towards a collaborative effort to address the problems of the global commons. Pause on AI will not solve this underlying problem. On the contrary, it may exacerbate it by diminishing technologies talismanic power.

The writer is Chairman, Center for Social and Economic Progress

The Indian Express (P) Ltd

First published on: 01-05-2023 at 07:44 IST

More here:

Vikram Mehta writes: Why we cant pause AI - The Indian Express

Have you heard of the term Artificial General Intelligence (AGI)? If not, let me clarify. AGI can be thought of as an AI system that can understand, process, and respond the intellectual tasks just like humans do. It's a challenging task that requires an in-depth understanding of how the human brain works so we can replicate it. However, the advent of ChatGPT has drawn immense interest from the research community to develop such systems. Microsoft has released one such key AI-powered system called HuggingGPT (Microsoft Jarvis). It is one of the most mind-blowing things that I have come across.

Before I dive into the details of what is new in HuggingGPT and how it works, let us first understand the issue with ChatGPT and why it struggles to solve complex AI tasks. Large Language models like ChatGPT excel at interpreting textual data and handling general tasks. However, they often struggle with specific tasks and may generate absurd responses. You might have encountered bogus replies from ChatGPT while solving complex mathematical problems. On the other side, we have expert AI models like Stable Diffusion, and DALL-E that have a deeper understanding of their subject area but struggle with the broader tasks. We cannot fully harness the potential of LLMs to solve challenging AI tasks unless we develop a connection between them and the Specialized AI models. This is what HuggingGPT did. It combined the strengths of both to create more efficient, accurate, and versatile AI systems.

According to a recent paper published by Microsoft, HuggingGPT leverages the power of LLMs by using it as a controller to connect them to various AI models in Machine Learning communities (HuggingFace). Rather than training the ChatGPT for various tasks, we enable it to use external tools for greater efficiency. HuggingFace is a website that provides numerous tools and resources for developers and researchers. It also has a wide variety of specialized and high-accuracy models. HuggingGPT uses these models for sophisticated AI tasks in different domains and modalities thereby achieving impressive results. It has similar multimodal capabilities to OPenAI GPT-4 when it comes to text and images. But, it also connected you to the Internet and you can provide an external web link to ask questions about it.

Suppose you want the model to generate an audio reading of the text written on an image. HuggingGPT will perform this task serially using the best-suited models. Firstly, it will generate the image from text and use its result for audio generation. You can check the response details in the image below. Simply Amazing!

HuggingGPT is a collaborative system that uses LLMs as an interface to send user requests to expert models. The complete process starting from the user prompt to the model till receiving the response can be broken down into the following discrete steps:

In this stage, HuggingGPT makes use of ChatGPT to understand the user prompt and then breaks down the query into small actionable tasks. It also determines the dependencies of these tasks and defines their execution sequence. HuggingGPT has four slots for task parsing i.e. task type, task ID, task dependencies, and task arguments. Chat logs between the HuggingGPT and the user are recorded and displayed on the screen that shows the history of the resources.

Based on the user context and the available models, HuggingGPT uses an in-context task-model assignment mechanism to select the most appropriate model for a particular task. According to this mechanism, the selection of a model is considered a single-choice problem and it initially filters out the model based on the type of the task. After that, the models are ranked based on the number of downloads as it is considered a reliable measure that reflects the quality of the model. Top-K models are selected based on this ranking. Here K is just a constant that reflects the number of models, for example, if it is set to 3 then it will select 3 models with the highest number of downloads.

Here the task is assigned to a specific model, it performs the inference on it and returns the result. To enhance the efficiency of this process, HuggingGPT can run different models at the same time as long as they dont need the same resources. For example, if I give a prompt to generate pictures of cats and dogs then separate models can run in parallel to execute this task. However, sometimes models may need the same resources which is why HuggingGPT maintains an attribute to keep the track of the resources. It ensures that the resources are being used effectively.

The final step involves generating the response to the user. Firstly, it integrates all the information from the previous stages and the inference results. The information is presented in a structured format. For example, if the prompt was to detect the number of lions in an image, it will draw the appropriate bounding boxes with detection probabilities. The LLM (ChatGPT) then uses this format and presents it in human-friendly language.

HuggingGPT is built on top of Hugging Face's state-of-the-art GPT-3.5 architecture, which is a deep neural network model that can generate natural language text. Here is how you can set it up on your local computer:

The default configuration requires Ubuntu 16.04 LTS, VRAM of at least 24GB, RAM of at least 12GB (minimal), 16GB (standard), or 80GB (full), and disk space of at least 284 GB. Additionally, you'll need 42GB of space for damo-vilab/text-to-video-ms-1.7b, 126GB for ControlNet, 66GB for stable-diffusion-v1-5, and 50GB for other resources. For the "lite" configuration, you'll only need Ubuntu 16.04 LTS.

First, replace the OpenAI Key and the Hugging Face Token in the server/configs/config.default.yaml file with your keys. Alternatively, you can put them in the environment variables OPENAI_API_KEY and HUGGINGFACE_ACCESS_TOKEN, respectively

Run the following commands:

For Server:

Now you can access Jarvis' services by sending HTTP requests to the Web API endpoints. Send a request to :

The requests should be in JSON format and should include a list of messages that represent the user's inputs.

For Web:

For CLI:

Setting up Jarvis using CLI is quite simple. Just run the command mentioned below:

For Gradio:

Gradio demo is also being hosted on Hugging Face Space. You can experiment with it after entering the OPENAI_API_KEY and HUGGINGFACE_ACCESS_TOKEN.

To run it locally:

Note: In case of any issue please refer to the official Github Repo.

HuggingGPT also has certain limitations that I want to highlight here. For instance, the efficiency of the system is a major bottleneck and during all the stages mentioned earlier, HuggingGPT requires multiple interactions with LLMs. These interactions can lead to degraded user experience and increased latency. Similarly, the maximum context length is also limited by the number of allowed tokens. Another problem is the System's reliability, as the LLMs may misinterpret the prompt and generate a wrong sequence of tasks which in turn affects the whole process. Nonetheless, it has significant potential to solve complex AI tasks and is an excellent advancement toward AGI. Let's see in which direction this research leads us too. Thats a wrap, feel free to express your views in the comment section below.Kanwal Mehreen is an aspiring software developer with a keen interest in data science and applications of AI in medicine. Kanwal was selected as the Google Generation Scholar 2022 for the APAC region. Kanwal loves to share technical knowledge by writing articles on trending topics, and is passionate about improving the representation of women in tech industry.

Continued here:

HuggingGPT: The Secret Weapon to Solve Complex AI Tasks - KDnuggets