Optimism, Concern, and Paranoia

Beginning in late 2022, the public availability of advanced Large Language Models (LLMs) and their chat-based applications (OpenAI’s ChatGPT, Meta’s LLaMa, Anthropic’s Claude), served as the spark that set off a powder keg of hype around Artificial Intelligence (AI) technologies that had been building up for a long time. Reactions have ranged from unbounded optimism about productivity capabilities to paranoia around machines surpassing human intelligence, leading to the outright destruction of the human race after a quick detour through a horrifying dystopian nightmare.

In reality, what is likely needed is an even-handed approach to adapt to the dizzying pace of innovation associated with AI, embracing the massive opportunity that it brings while readying society and the labor force for the coming changes. This is not the first time that a set of technologies has come along and disrupted many sectors at once. Although things are likely to move much more quickly this time, we can look to history for some examples of how (and how not) to prepare.

Historical Echoes on a New Digital Frontier

The question, “What will happen to all the horses?” alludes to the early 20th-century quandary when the internal combustion engine was new and the burgeoning automobile industry was rendering horse-driven transport methods obsolete. The fear then was palpable. Workers worried about job losses, society was anxious about safety, and governments struggled with the question of how to regulate automobiles and their effects on everything from city planning to streetcar manufacturing. The current AI wave feels eerily similar in some aspects of the scope and scale of anticipated disruption.

AI technologies, especially generative AI, are creating a plethora of opportunities. There is justification for substantial optimism if only because of the sheer scope of problems that AI can address. As Andrew Ng, cofounder and one-time head of Google Brain, famously said in 2017, “AI is the new electricity. Just as 100 years ago electricity transformed industry after industry, AI will now do the same.”

We would be wise to find ways to embrace as much of that opportunity as we can while maintaining agility in our ability to do so responsibly.

This is no easy feat, and with every stride forward that AI takes, concerns deepen in some circles about its implications.

Societal, Economic and Ethical Concerns

While LLMs and associated applications are capable of drafting emails or assisting with customer support, what becomes of the assistants and support staff?

Just as carriage drivers did with the advent of the automobile, traditional jobs will fade in the face of automation. However, history has shown that technology also creates new jobs, even if it means that the workforce must adapt and adjust. This workforce disruption will not be all negative. Immediate positive changes will also be felt in the day-to-day lives of many office workers, for example, who can refocus the effort they are currently expending on procedural tasks to focus on activities with a higher value add.

The influence AI will have on our daily lives, decision-making processes, and cultural values will also be profound. As these models become more integrated into our lives, they will shape our perceptions, sometimes without us even realizing it’s happening. In fact, it’s already happening as illustrated by the wildfire-like way that misinformation spreads via social media and other internet-powered routes. Suggestion engines that have the sole purpose of promoting content that gets more clicks are nothing new. With generative AI, the quantity and reach of that content and its ability to convince and persuade people could grow exponentially. This is but one immediate example of why a healthy amount of concern around the use of these technologies and urgency in finding ways to ensure that they are used responsibly are also justified.

Top AI ML News: The Human Brain is the Best Example of a Supercomputer

Navigating the Future

For organizations and individuals, navigating the AI revolution means understanding its nuances, embracing its advantages, and being vigilant about its risks.

1. Education: Equip the current and upcoming workforce with AI literacy, which simply means a basic understanding of what these technologies can and cannot do, and a surface understanding of how they do it. This will be as crucial as basic digital literacy has been since the turn of this century.
2. Ethical AI: Organizations must adopt principles and policies that ensure their AI systems are transparent, explainable, and, to the extent possible, devoid of harmful biases. Helpful tools, such as the AI model documentation method known as model cards, already exist and can provide a start on the transparency and accountability that will be essential to the responsible use of AI.
3. Regulation and Oversight: Governments and other regulatory bodies will play a pivotal role in shaping the ethical trajectory of AI development. Thoughtful regulation should ensure that societal interests are protected while also allowing innovation to thrive.
4. Human-AI Collaboration: Rather than viewing AI as a competitor, we should focus on collaborative models of working with these technologies. AI can handle data-driven tasks while humans bring creativity, emotional intelligence, and context.

For years, science fiction has painted images of dystopian futures where AI surpasses human intelligence, leading to unfavorable outcomes. These narratives tap into a very real human fear: loss of control. But the coming AI revolution is as much an opportunity as it is a challenge. Here in the real world, AI should be approached as a tool, not a replacement.

The future may seem daunting from our current vantage point, but it is worth remembering that every significant technological shift has been met with similar trepidation.

Proper guidelines, sensible oversight, and enlightened ethical frameworks surrounding AI’s use can ensure that AI serves humanity rather than subjugating it. After all, none of us wants to be the one still asking 30 years from now, “What will happen to all the horses?” as we get hit by a bus.

[To share your insights with us, please write to sghosh@martechseries.com]

The post “What Will Happen to All the Horses?” – Surviving the Coming AI Revolution appeared first on AiThority.

What does the future look like for generative AI? If hallucinations persist, it’s a complicated question.

Conversations about the transformative possibilities of generative AI tools invariably come with a cautionary note about hallucinations, i.e., the tendency for AI tools to conjure fabricated information seemingly out of nowhere. AI experts have differing opinions about how long it will take for the industry to get hallucinations under control — or if it’s even possible to eliminate them.

In any case, AI hallucinations will continue to pose problems for the foreseeable future. That means businesses must learn to spot and work around these hallucinations if they want to leverage the potential benefits of generative AI tools in their daily operations. And, it begins with understanding why AI hallucinations occur and what employees can do to limit their frequency and reach.

The Origins and Implications of Generative AI Hallucinations

In May 2023, an attorney representing an injured airline passenger submitted a collection of cases as part of a legal brief.

The problem? Six of the cases didn’t exist.

During the research process, the lawyer used ChatGPT, which hallucinated the existence of certain cases.

How does an error like this happen?

While AI experts don’t fully understand what causes AI hallucinations, unfocused, massive sets of training data — coupled with the complexity of the generative process itself — is a primary suspect.

Essentially, large language models (LLMs) like ChatGPT are trained on a large set of sometimes outdated training data. Because most LLMs can only draw conclusions based on the user patterns displayed within the training data, unfamiliar situations can pressure the tools into making false or misleading claims. The more gaps or inconsistencies in the training data, the higher the likelihood of a hallucination.
When certain LLMs pull from sources outside of their initial training data, it creates a new set of challenges. For example, consider the amount of disinformation, parody, and bias that exists on the internet. Even humans struggle to determine what’s real and what’s fake (in part thanks to AI tools themselves). LLMs must navigate this labyrinth of information to produce an accurate output — without the familiarity of internet culture and jargon that humans possess.

This complexity directly contributes to hallucinations, placing a clear limitation on generative AI’s usability in certain industries — as our previous example with the attorney illustrates. Other high-stakes industries like healthcare are experimenting with how generative AI can be used for administrative tasks. However, more substantive generative AI use cases remain out of reach until the industry can get a handle on the hallucination problem.

How to Work Around AI Hallucinations

While generative AI hallucinations may prove difficult to eradicate entirely, businesses can learn to minimize their frequency. But, it requires a concerted effort and industry-wide knowledge sharing.

Here are three tactics you can use right now to mitigate the impact of generative AI hallucinations in your work.

• Keep humans in the driver’s seat

Think of generative AI like GPS. You wouldn’t blindly follow GPS instructions if they told you to drive off of a cliff, and you should never take generative AI’s outputs as gospel or use them as the sole basis for decision-making.

Instead, treat generative AI as a supplemental tool and encourage your employees to double-check any information these tools produce. Emphasize that you’re not introducing these tools to replace employees, but to make their lives easier. Cultivating an environment in which employees share generative AI best practices with one another and remain aware of new developments provides a foundation for your organization to stay one step ahead of hallucinations.

• Prioritize prompt refinement

With generative AI, the more specific your prompts, the better.

For example, suppose you want to learn about the history of cloud computing. You log into ChatGPT and type, “Tell me about the history of cloud computing.” The model produces a massive wall of text that is overwhelming to sift through. This shouldn’t be a surprise. General or vague prompts often produce vague answers — and this ambiguity is where hallucinations can sneak in undetected.

A better approach is to narrow your prompts and include relevant details. To round out our cloud computing example, here are a few refined prompts that should reduce the likelihood of hallucinations:

• What were the three defining events that contributed to the rise of cloud computing? Keep your answer to two paragraphs or shorter.
• You are a historian recounting the origins of cloud computing technology. In your account, exclude any mentions of modern cloud computing.
• Tell me about the history of cloud computing — specifically about the role DARPA played in the technology’s evolution.

This level of specificity and instruction will help you receive much more detailed, focused responses and make it easier to spot potential hallucinations when they occur.

• Consider a purpose-built LLM

While tools like ChatGPT and Bard have obvious value, your business could potentially achieve stronger results from building your own purpose-built LLM.

A purpose-built LLM provides a narrow focus, using a smaller set of training data to deliver tailored responses. The biggest advantage of this model is the ability to incorporate your organization’s own internal data into a controlled dataset and solve problems unique to your business and customers. This level of control limits the likelihood of hallucinations — as long as you keep the scope of your prompts narrow.

Ultimately, I think we’ll see these more personalized applications of generative AI grow in popularity for businesses. They may not attract the media attention of a ChatGPT, but these models have more practical applications for organizations without the heightened risk of hallucinations that come with larger tools.

The Unclear Future of Generative AI Hallucinations

There’s no way around it: Generative AI hallucinations will continue to be a problem, especially for the largest, most ambitious LLM projects.

Though we expect the hallucination problem to course correct in the years ahead, your organization can’t wait idly for that day to arrive. To reap the benefits of generative AI now, you need to understand how to prevent these hallucinations and flag them when they occur, whether they pop up in ChatGPT or your own purpose-built LLM.

[To share your insights with us, please write to sghosh@martechseries.com]

The post Is Generative AI’s Hallucination Problem Fixable? appeared first on AiThority.

The recent research in the field of computational neuroscience have shown a remarkable evolutionary relationship between cognitive features of a human brain and its computing prowess. The human brain is actually wired to perform like a supercomputer. A good number of brain activities are actually happening in a specific way that can be closely related to advanced mathematical models for visual information processing. According to the scientists, our brain functions on the principle of Bayesian Inference, a model that very much operates the world’s most advanced supercomputers.

According to a published study, scientists* have developed an empirical mathematical model for neural decoding, closely matching the functions of a human brain when inferring environmental attributes using image data. This model simulates neural communication carried out using Bayesian inference. This statistical technique combines sensory data acquired from previous interactions with the environment and use new observations to make an intelligent assessment of the present situation. For example, if you see a small black animal with a snout and four legs with a tail near a drain or street, you can guess, it could be a mice. This sensory inference helps us differentiate between animals, birds, reptiles, trees, and non-living things.

The study’s senior investigator Dr Reuben Rideaux, from the University of Sydney’s School of Psychology, said: “Despite the conceptual appeal and explanatory power of the Bayesian approach, how the brain calculates probabilities is largely mysterious.”

“Our new study sheds light on this mystery. We discovered that the basic structure and connections within our brain’s visual system are set up in a way that allows it to perform Bayesian inference on the sensory data it receives.

“What makes this finding significant is the confirmation that our brains have an inherent design that allows this advanced form of processing, enabling us to interpret our surroundings more effectively.”

The study’s findings not only confirm existing theories about the brain’s use of Bayesian-like inference but open doors to new research and innovation, where the brain’s natural ability for Bayesian inference can be harnessed for practical applications that benefit society.

“Our research, while primarily focused on visual perception, holds broader implications across the spectrum of neuroscience and psychology,” Dr Rideaux said.

“By understanding the fundamental mechanisms that the brain uses to process and interpret sensory data, we can pave the way for advancements in fields ranging from artificial intelligence, where mimicking such brain functions can revolutionize machine learning, to clinical neurology, potentially offering new strategies for therapeutic interventions in the future.”

The research team, led by Dr William Harrison, made the discovery by recording brain activity from volunteers while they passively viewed displays, engineered to elicit specific neural signals related to visual processing. They then devised mathematical models to compare a spectrum of competing hypotheses about how the human brain perceives vision.

Scientists:

William J. Harrison, Paul M. Bays & Reuben Rideaux

Source: The University of Sydney

[To share your insights with us, please write to sghosh@martechseries.com]

The post The Human Brain is the Best Example of a Supercomputer appeared first on AiThority.

While artificial intelligence is ginormous and revolutionary, generative AI, let’s say is evolutionary and more promising. Salesforce‘s research team is constantly seeking new methodologies to solve problems across different societies, and society at large. This article is an attempt to understand how Salesforce’s partnership with an academic institution and a biomedical company can apply the AI language model.

On a global level, leaders across industries have whole-heartedly acknowledged and appreciated the possibilities Generative AI can bring to a company. The year 2022, can be christened as the year of Generative AI – when technology took the baton and created novel content in different formats and in a perfectly human-like tone.

Recommended: Marketing Automation: A Complete Guide

The Works of Nikhil Naik & the Salesforce Research Team

Generative AI is a hot-selling topic, courtesy, of innovative products like ChatGPT, which have been in the news ever since its launch in November 2022. Over the past few months, multiple articles have been written about its superpowers while some raised concerns over its ethical usage.

Be it writing songs, jokes, essays, articles, etc seems lucrative right now, but what’s intriguing and most likely will yield results is the work Salesforce’s research team headed by Nikhil Naik, Director Of AI Research, is doing. In the last 5 years, Naik’s team has been quietly and passionately working on some bigger applications.

The research team trained generative AI on conversational language, which was then shaped into a development code through a large-scale language model known as CodeGen.

AI for Society Initiative

Naik explained that the goal of the research team under the AI for Society initiative was simple – to be able to apply Salesforce AI on issues that are likely to have a bigger impact on society.

Through this initiative, the Salesforce Research work has been quite vivid ranging from incorporating computer vision to track great white sharks and identifying the correct treatment plan for breast cancer patients via artificial intelligence to implementing Ai for balanced economic policies.

Naik added that the team had an interesting way of working. They first identify the AI tools and techniques they excel in and then look out for problems where those AI techniques would work in a tailor-made fashion. This unique approach resulted in the creation of ProGen – an AI language model trained on the world’s largest protein database.

Recommended: Surviving the Digital Marketing Storm -Navigating Recession, Privacy and Disruptive Innovation

Naik pointed out,

“AI models ingest a large amount of text and they learn to predict the next word that might come after a given word. And just by training using this pretty simple method, you can train an AI algorithm to generate very realistic language about any topic that you might be interested in. And what we realized is that the same technology can be applied to generating proteins.”

When an opportunity of developing novel proteins came through, the Salesforce Research team looked at the avenues it would open up like new medicines, vaccines, or sustainability innovations.

In 2020, Naik and his troop set out to work around the problem of protein design with the help of generative AI, especially large language models. The reason for choosing protein design was simple – it was a vast field with tremendous scope for invention and research, which meant, “accelerating the discovery of novel drugs and useful industrial chemicals.”

An Alphabet Using Amino Acids

At first, the Salesforce team went all creative and made an “alphabet” using nothing but amino acids. After all, what better way to build than using the building block of all proteins? Just the way “letters” come together to form proteins, similarly, you can train a large language model to not only predict the next word but also generate sentences in English. The team trained used a database of 280 million protein sequences to generate novel proteins and train a large language model.

Despite their excitement about the progress, Naik and the team did not have the bandwidth to test and ascertain if the AI language model for generating proteins could create something useful. To fight ambiguity, they partnered with the Fraser Lab at UCSF and medical startup Tierra Biosciences to examine their research.

First, the Salesforce Research team sent around 100 AI-generated proteins to synthesize to Tierra and create test tube versions of them. Tierra found that the proteins were functional, and so for further research, they were sent to the University of California San Francisco’s Fraser Lab. The lab drew comparisons between natural proteins and artificial proteins.

“The lab tests showed that we can design proteins that are 60-70% dissimilar to anything ever seen in nature, but that are still functioning proteins, containing biological activity. And that is an important scientific milestone for the future of drug discovery and industrial chemical design,” Naik said.

Another striking conclusion was that ‘ProGen-created proteins were 73% biologically active, whereas only 59% of naturally-occurring proteins were.’

Recommended: How Agencies Can Use AI Image Generation to Build Better and More Creative Ads

Using Technology Ethically

Having understood the intricacies of the work, Naik and his team, and practically anyone who was even remotely involved, knew that using this powerful technology ethically was of paramount importance.

Each and every step and process were thoroughly scrutinized by Salesforce’s Ethical AI Council and Office of Ethics under Salesforce’s Chief Ethical and Humane Use Officer, Paula Goldman to ensure ethical guidelines were followed and that Ai was deployed securely and responsibly.

Considering the limitless opportunities in this space and to further strengthen the ethical usage of the technology, Goldman and her teams decided to build on Trusted AI Principles to help guide the process. This was more critical in the case of ProGen, where there was a strong need for protocols to be put in place to “ensure safe usage and limitation of unintended harmful effects.”

Nikhil and his team are optimistic, and excited, yet modest about their ability to create new designs of proteins never-before-seen in nature, because of the fact that they can be used for medicine and other domains.

And the Journey Continues

Ever since the success of the mentioned experiments, researchers have been more than keen to build on his team’s work and also showcase the applications in different domains. With a hopeful wish, Nikhil stated that in the near future, there will most likely be a massive spike in research and commercial activity in this space. The journey seems to have started with Naik and his Salesforce AI Research team making every effort to identify potential treatments for rheumatoid arthritis, multiple sclerosis, and other neurological and autoimmune disorders by making the most of their work with ProGen.

What Naik and his team are doing earnestly, and the kind of approach they have is truly remarkable in every possible way. The question remains, can it be used somewhere else, in some other field or to address global challenges like food supply, sustainability, or climate change? Naik proudly says that this is where we can see unlock the true power of AI. As they say, the journey has just begun.

Story input: Salesforce

The post From Creating Alphabet’s to Developing Novel Proteins: Salesforce’s Generative AI To Tackle Society’s Biggest Challenges appeared first on AiThority.

Alto Neuroscience reported results from its Phase 2a study of ALTO-100, demonstrating clear evidence of efficacy and favorable safety in patients with MDD. In the study, patients with a biomarker profile that ties back to a mechanistic understanding of ALTO-100 and depression exhibited a significantly greater change in Montgomery–Åsberg Depression Rating Scale (MADRS) scores and response rates than those without the biomarker profile. This first-of-its-kind Phase 2 study leveraged Alto’s Precision Psychiatry Platform to identify likely drug responders based on an understanding of biological heterogeneity in depression and ALTO-100’s novel mechanism. The results support the initiation of ALTO-100 into a large Phase 2b trial in patients with biomarker-defined MDD in January 2023. Topline data from the Phase 2b study is anticipated to readout in the first quarter of 2024.

Artificial Intelligence News : Merlin AI, the Evolution of ChatGPT is Now a Freestanding Mobile App

“The strength of these results demonstrates, for the first time, that we can prospectively identify likely responders to our novel drugs and apply data-driven measurement to the treatment of psychiatric and other central nervous system disorders”

Topline results from the ALTO-100 Phase 2a study include:

• The biomarker-defined MDD patient group (n=59) demonstrated a 15.5-point mean MADRS reduction compared to 10.6 points in the patient group without the biomarker profile (n=64) at week 6 (p=0.001, d=0.6).
  • In patients taking ALTO-100 as monotherapy (n=45), a 17.4-point change in MADRS was observed at week 6 in the biomarker-defined group compared to 11.8 points in the group without the biomarker (p=0.026, d=0.66).
  • In patients taking ALTO-100 in addition to another antidepressant (i.e., as adjunctive treatment) (n=78), a 14.4-point change in MADRS was observed at week 6 in the biomarker-defined group compared to 9.9 points in the group without the biomarker (p=0.013, d=0.56).

• 61% of biomarker-defined patients achieved clinical response (defined as ³50% reduction in depression symptoms) compared to 33% of patients without the biomarker profile (p=0.004).
  • 81% of biomarker-defined patients taking ALTO-100 as monotherapy achieved clinical response compared to 38% of patients without the biomarker profile (p=0.01).
  • 50% of biomarker-defined patients taking ALTO-100 as adjunctive treatment, achieved clinical response compared to 31% of patients without the biomarker profile (p=0.07).

• On the Clinician Global Impression of Severity (CGI-S) scale, biomarker-defined patients demonstrated an improvement of 1.66 points (5-point scale) from baseline to week 8 compared to an improvement of 1.17 points in patients without the biomarker profile (p=0.02, d=0.41).
• ALTO-100 has now been studied in more than 395 subjects and has displayed a favorable tolerability profile. No new safety signals were observed in this study.

“The strength of these results demonstrates, for the first time, that we can prospectively identify likely responders to our novel drugs and apply data-driven measurement to the treatment of psychiatric and other central nervous system disorders,” said Amit Etkin, M.D., Ph.D., founder and chief executive officer of Alto Neuroscience. “We are pleased to have finished this study ahead of schedule, and I am proud of our team for their relentless persistence and execution to reach this important milestone. As this provides a substantial level of de-risking, we are eager to move ALTO-100 into the Phase 2b study which will begin enrollment this month.”

Artificial Intelligence Technology : How Artificial Intelligence Will Help Facilitate the Hybrid Working Environment

The study was an 8-week open-label, holdout dataset-controlled study to evaluate the efficacy and safety of ALTO-100 in patients with MDD or post-traumatic stress disorder (PTSD). 133 patients with primary MDD and 95 patients with primary PTSD were enrolled in the study. Alto utilizes a rigorous data science approach and prospective replication to predict clinical efficacy in holdout datasets and define reliable drug predictors while avoiding false discovery. The primary endpoint was change from baseline in MADRS score at week 6, with a replication threshold pre-specified as a Cohen’s d effect of 0.5 or greater.

Adam Savitz, M.D., Ph.D., chief medical officer of Alto Neuroscience, added, “We are encouraged by the potential of these results to redefine the treatment paradigm in depression, which today is largely dependent upon trial-and-error. Using our biomarker platform and data analytic approach, we’ve demonstrated a clear clinical signal for ALTO-100 in a particularly underserved patient population. Stronger clinical signals in the MDD population defined by our biomarker profile support movement into a larger Phase 2b study in this population. Analyses are ongoing for the population with PTSD to inform potential future studies in that indication. We look forward to building on the learnings from this study for additional medications and patient subgroups as we continue to validate our precision psychiatry approach.”

 AI in Advertising : AI in Advertising and Marketing – Hype, Disillusionment, Back to Hype Again?

 [To share your insights with us, please write to sghosh@martechseries.com] 

The post Alto Neuroscience Announces Positive Results for ALTO-100 in Phase 2 Study Supporting Advancement of First-in-Class Mechanism for Treating Depression appeared first on AiThority.

MindMaze, a global pioneer in the development of neurotechnology, confirms a new partnership with Alfa Romeo F1 Team ORLEN. Through this partnership, MindMaze will expand its groundbreaking research using its MindDrive brain technology platform to bring together advanced neuroscience, state-of-the-art technology, and engineering to boost safety and human performance in motorsport.

Top Artificial Intelligence Insights: Determining the Potential of Your AI Algorithm Starts with Measurement

“We are delighted to be working closely with the Alfa Romeo F1 team to expand our partnership. Human performance is tested to the maximum in F1, which is why the research we are conducting with team ORLEN is so important. Our goal is not only to generate and study data that benefit human performance in motorsport, but also to inform the future of innovation for universal safety features across the entire automotive industry,” says MindMaze CEO Tej Tadi.

MindDrive’s brain technology platform is part of the company’s R&D innovation division— MindMaze Labs. The collaboration will include holistic research assessing both car and driver. The brain technology platform will specifically monitor human performance by capturing brain data, both on and off the racetrack. MindMaze’s F1 partnership represents a novel new field of study that underscores the company’s objective to transform brain health by decoding the brain and harnessing its ability to accelerate recovery from injury, learn, and adapt.

Latest Aithority Insights: Why Contextual Targeting Deserves Another Look with Artificial Intelligence (AI)

Frédéric Vasseur, Team Principal of Alfa Romeo F1 Team ORLEN, added, “Innovation is at the heart of everything we do in Formula One, so I’m proud to welcome MindMaze as a neurotechnology pioneer to our partner family. It is a company that has shown its research has real-world benefits, and I’m looking forward to combining our deep technical competencies to see what insights we can unlock together.”

MindMaze recently showcased the brain technology and its motorsport research at the Miami Grand Prix, where MindMaze was a founding partner. As part of the agreement, MindMaze will have brand sponsorship on the safety helmets of Alfa Romeo F1 Team ORLEN drivers Valtteri Bottas and Zhou Guanyu for the remainder of the 2022 F1 season.

MindMaze’s F1 programme is part of its growing portfolio of research in motorsport. Currently, it has a MindDrive-focused research project underway in the US with the Andretti Autosport Indycar team and its brain health ambassador, Romain Grosjean. Previous partners include F1’s McLaren and Haas teams.

AI and ML News: Why SMBs Shouldn’t Be Afraid of Artificial Intelligence (AI)

[To share your insights with us, please write to sghosh@martechseries.com]

The post MindMaze Enters New Strategic Partnership with Alfa Romeo F1 Team ORLEN appeared first on AiThority.

The post Inscopix Announces Cloud-Based IDEAS Platform for Neuroscience Research with New Neural Circuit Workflow appeared first on AiThority.

The post Inscopix Launches New Integrated Miniscope-Behavioral Platform and Co-marketing Collaboration with Noldus appeared first on AiThority.

Every year, millions of children between the age groups of five and fourteen years are diagnosed with a peculiar kind of neuro-biological development disorder called Autism Spectrum Disorder (ASD). In 2016, 62 million cases of ASD were reported globally. Despite impressive neuroscience innovations in the medical field in the last five decades, scientist have been unable to find a treatment for ASD. However, an early diagnosis of ASD can improve the lives of children suffering from autism. Now, researchers have found the benefits of machine learning algorithms in early detection of ASD and how different neuroscience techniques could be used to identify speech patterns among children with development problems related to communication, sensory processing and common social interactions.

The latest study was conducted by a group of experts in neuroscience, data curation, formal analysis and data visualization. The National Institutes of Health, Health and Medical Research Fund (Hong Kong: 02130846, PI: PW), Global Parent Child Resource Centre Limited and Dr Stanley Ho Medical Development Foundation provided grants to support this research on the role of machine learning in identifying the cross-linguistic patterns of speech prosodic differences in autism.

Recommended AI ML News: SpectrumAi Announces $9 Million in Seed Funding to Innovate Autism Care

Individuals with Autism Spectrum Disorder or Autism Spectrum Condition (ASD / ASC) show a wide range of characteristics that are linked to speech prosody. However, there was no common ground of research on ASD and how it varies among individuals who speak non-English languages. The researchers from the Northwestern University in the US collaborated with others based in Hong Kong to identify the differences between individuals who speak either English or Cantonese or both. The differences in the speech prosody among children with ASD was studied using supervised machine learning analytic approach to encode changes in loudness or pitch, emotions, and also underlying genetic influences on speech prosody.

Autism in US and How Machine Learning is Marching Forward

According to the Centers for Disease Control and Prevention, autism prevalence in the United States has risen significantly in recent years, from 1-in-150 in 2000 to 1-in-44 in 2022. Diagnosis rates among female and minority populations continue to lag, but novel telepsychiatry and automated diagnostic solutions are beginning to bridge the diagnosis gap.

While the U.S. is the global leader in ABA therapy, critical provider shortages still exist nationally, with more than half of all U.S. counties registering zero supervisory Board-certified Behavior Analysts, and 49 states not reaching the per capita benchmark supply of certified ABA providers. All 50 states now mandate ABA coverage for up to 40 hours per week which has led to private equity- and venture capital-backed ABA organization consolidation and growth to meet consumer demand that massively outpaces the current supply 10-to-1.

Machine learning could be used to identify how speakers with ASD display varying pitch, in addition to having a slower speech rate and oddness of stress on syllables. By applying multivariate techniques of machine learning on acoustics features such as ‘rhythm’ and ‘intonation’, researchers were able to successfully classify individuals with “typical development” or TD from those who showed ASD features. ML Classification, based on Linear Support Vector Machines (SVMs), of TD versus ASD population was further used to ascertain the diagnostic characteristics of cognitive, neurological and speech impairments.

Why researchers used SVM in this ASD study?

The research paper fairly outlined the logic behind the use of support vector machines for ML classification of ASD features. It said, “the linear SVM is preferable to non-linear kernels because theoretically it is always possible to find a linear decision boundary that separates data, in spite of high data dimensionality and small sample size.”

Also, SVM is perfect for handling of high dimension data that are often linked with the various types of speech acoustic features linked to ASD.

Future of Neuroscience to Correct ASD

Machine learning will play a big role in the ASD and its management. From gene expression to neuro-imaging to fundamentally understanding a speaker’s progress as a TD individual, machine learning’s role transcends culture and hereditary.

[To share your insights with us, please write to sghosh@martechseries.com]

The post Researchers Find Innovative Use of Machine Learning in Diagnosing Autism Spectrum Disorder (ASD) appeared first on AiThority.

Hi Samantha, please tell us about your current role and how you began to work with AI and data science.

I’m currently an Associate Professor of Computer Science at Stevens Institute of Technology, where I lead the Health and AI Lab. I’ve always combined computing and biology in some form, starting with biomathematics and looking at properties of DNA, then bioinformatics and gene expression, and finally biomedical informatics working on electronic health record data. Now I’m focused on how we can support both doctors and patients making health decisions, whether that means using AI to give them information they otherwise wouldn’t have, or to figure out what the right information is.

Since you first started with AI, how much has the technology and the applications evolved for the general consumers and researchers?

When I first started AI for health was not widespread, and the data was mainly clinical data collected by a few hospitals. Now it has grown into its own field, with dedicated conferences, shared datasets, and real-world success stories of AI informing and improving medical care. A particularly exciting area of growth is the data being collected by people outside medical settings using wearable devices and apps. We now have much more context about people’s health and choices during daily life, rather than only during medical encounters. This is now enabling AI that can help people make daily decisions.

Recommended: AiThority Interview with Robert Schena, CEO and Co-founder at Rajant

Please tell us about Steven’s Health and AI Lab. What is the idea behind putting AI at the center of human healthcare management?

I put humans at the center of health, and always aim to develop systems that can fulfill human needs and work with people to achieve a goal, rather than trying to replace them. I think AI has a lot of potential as an interface between data and people. It’s hard to make sense of the thousands of blood glucose measurements someone may collect and to use that to inform how they manage their diabetes, but AI can help find why blood glucose changes and help inform behavior, like adjusting insulin dosing the day after exercise.

What kind of outcomes are you hoping to achieve with AI Lab?

We aim to generate new knowledge about health (like why some people regain consciousness after a stroke while others don’t) and use it to improve decisions (like how a patient in a hospital is cared for, or how people manage their chronic conditions or maintain their health).

There are a lot of things that can influence food choices — but AI probably wouldn’t be one of them! Could you please tell us how AI is being used to make people turn to healthier diets?

AI is already influencing our food choices in ways we often don’t think about. It determines what food ads we see when we look at a website, and what restaurants show up at the top of search results. But traditionally what we know about what we should eat comes from generic guidelines. The biggest recent change has been the use of AI to provide personalized nutrition guidance, so figuring out what kinds of foods will keep your blood sugar healthy. To get this information though you either need to be part of a research study or pay a company, so it’s not very accessible. The new NIH Nutrition for Precision Health program aims to change this by learning not just learning individualized relationships between diet and health but doing this in a way that we can provide information to people who aren’t part of the study. The AI part is taking the huge amount of data collected (microbiome, genome, social determinants of health, what foods people eat, various indicators of health etc.) and learning the relationship between these factors.

Recommended: AiThority Interview with Pete Stein, President at Merkle Americas

Do you agree the same AI algorithms could be used to influence / intimidate people into buying unhealthy foods? What is the regulation on these applications in the long run, particularly when vulnerable classes and minor populations are involved?

I work mainly on causal inference, so trying to understand what leads to changes in blood glucose, or what puts people at risk of gestational diabetes. A major advantage of this kind of AI, compared to predictive models, is that people can understand the recommendations, and see why an algorithm thinks they should take an action. This makes it easier to know when to trust it.

On the other hand, while there’s oversight of academic research with people, there is not necessarily the same ethics review of algorithms. If an app or algorithm for decision support isn’t regulated by the FDA, there’s no other oversight right now. The FDA is in the process of updating its guidance so we may see more regulation of algorithms that currently are not under their purview.

Recommended: AiThority Interview with Adrian Jones, Chief Revenue Officer at ABBYY

A message to every AI professional looking to build a career in food and healthcare industry.

I think the biggest thing AI professionals could do but don’t always do is talk to and try to deeply understand users – what they want AI to do, how they see it fitting into their lives, what concerns they have.

Thank you, Samantha! That was fun and we hope to see you back on AiThority.com soon.

[To share your insights with us, please write to sghosh@martechseries.com]

• About Samantha
• About Stevens Institute of Technology

About Samantha

Samantha Kleinberg is an Associate Professor in the Computer Science department at Stevens Institute of Technology.

After completing her PhD in Computer Science in 2010 from NYU, Dr. Kleinberg spent two years as a postdoctoral Computing Innovation Fellow at Columbia University in the Department of Biomedical Informatics. Previously, she was an undergraduate student at NYU in Computer Science and Physics. More recently, Dr. Kleinberg spent a year on sabbatical in the psychology department of University College London.

Dr. Kleinberg has written an academic book, Causality, Probability, and Time, and another for a wider audience, Why: A Guide To Finding and Using Causes. She’s the editor of Time and Causality Across the Sciences.

About Stevens Institute of Technology

Stevens Institute of Technology is a premier, private research university situated in Hoboken, New Jersey. Since our founding in 1870, technological innovation has been the hallmark of Stevens’ education and research. Within the university’s three schools and one college, 8,000 undergraduate and graduate students collaborate closely with faculty in an interdisciplinary, student-centric, entrepreneurial environment. Academic and research programs spanning business, computing, engineering, the arts and other disciplines actively advance the frontiers of science and leverage technology to confront our most pressing global challenges. The university continues to be consistently ranked among the nation’s leaders in career services, post-graduation salaries of alumni, and return on tuition investment.

The post AiThority Interview with Samantha Kleinberg, Assistant Professor at Stevens Institute of Technology appeared first on AiThority.