#ML tool

[bhaskarlive]

IIT Delhi researchers develop ML tool to map extent of landslides

Researchers from the IIT Delhi on Monday announced the development of a ground-breaking cloud computing and machine learning-based tool that can help map a landslide’s extent using satellite data.

This easy-to-use and publicly-available tool ML-CASCADE requires an approximate date and location of a landslide event.

Source: bhaskarlive.in

[rosekasa]

the major arcana

mum taught me how to read tarot cards when i was younger... ive always loved this card and it reminds me so much of ladynoir ;-;

hd and process

[katydoodles]

Thank you Ladynoir-Fan for the Kofi!

[etanow]

Something don't feel right

[anna-scribbles]

thirteen update 💕💍🍽️🩸

chapter 5: february

summary:

“These things do not concern you,” Papa told him flatly. “I will run my household however I see fit. Your concerns are with your schoolwork and your modeling.” Blood pumped heavy and fast through Adrien’s heart. That wasn’t—fair. Concern was about all he was capable of these days. “And what about Maman?” Adrien asked, exhausted, reckless. “May I be concerned about Maman?” Something shifted on Papa’s face, all his emotions smothered in stone.

excerpt:

The best day of Adrien’s life was eight months and six days ago. No contest.

It was a crisp kind of cold that day, the Paris sky blooming a bright and brilliant blue overhead. The sun pierced right through the brisk February air, a shock of spearmint and adrenaline in his veins. He couldn’t stop widening his eyes, couldn’t stop smiling. The city was so alive. Strains of love songs poured out of open cafe doors and onto tourists, their hands full of red roses and lovers’ hands. The cobblestones sang with the patters of paired footsteps all down the street. It was the city of love always, but today especially. Today Adrien was made of the stuff, just bursting with it.

And, like every other day in the running for the best of his life, Marinette was there.

“You’d better not pull anything,” she warned, tightening her grip on his hand as they passed by a tourist couple looking very… engrossed with each other in the middle of the street. “And—and if you do, you have to tell me. Right now.”

Marinette’s brow was lightly furrowed, the bridge of her nose just barely scrunched up. Her hair was pulled half-back with a pink ribbon, matching the shade of the skirt she wore beneath her velvety black peacoat. Her Mary Janes clipped anxiously down the road and Adrien’s heart danced and swelled and spun in his chest.

“Pull something? Me?” Adrien stepped aside so their arms were outstretched, and then pulled at Marinette’s fingers, sending her tumbling back into his arms. She looked up at him, trying to frown, smiling. He grinned. “I would never.”

���I’m serious.” Marinette untangled herself from his arms and interlocked her fingers again with his. Her hand was the warmest thing in the world. She looked at him sternly, wagging a finger in his face. “I need to know so I can—prepare. Especially if it’s something crazy. No funny business.”

“Marinette,” he moaned, draping a wounded hand over his heart. One corner of his mouth quirked into a smile, eyes darting to meet her gaze. “You think I’m funny?”

She groaned. “I think you‘re—I think you’re ridiculous, and sappy, and romantic, and I think it’s Valentine’s Day in Paris”—this part she shouted, which drew a few stares—“and I think you’re about to take me on an insanely adorable date, and I think Alya took me to get my nails done last week—!”

“You’re so thoughtful,” Adrien remarked, swinging their hands back and forth. “And observant. What a beautiful mind you have, my lady.”

“You have to tell me,” Marinette insisted. She stopped them on the street and frowned at him, pink flushing the apples of her cheeks. “Is it—are you—?”

“Hm?” Adrien murmured, brushing a lock of hair behind her ear. Marinette’s cheeks went ablaze.

“I—you—you know what I mean!” she spluttered. “Are you gonna…you know!”

He tilted his head to the side. “Am I…?”

read on ao3

[zoe-oneesama]

Saw your post about Andre

What do u think about Zoe, then?

As a character, she's not bringing anything worth getting excited about. Personality wise, she's just Marinette-Pre-Origins: sweet and eager for friends, but ultimately letting her fear of Chloe rule her behavior until someone else reaches out, and now she suddenly has a spine. And I'm all for the idea of Marinette taking the lessons she learned in Origins about standing up for herself and not sitting by and letting injustices happen, and then turning around and inspiring someone else to do the same, so on paper, Marinette being that person for Zoe would be sweet and make sense. Except we have a full class of kids that she could've and should've already been that inspiration for, so this feel very Too Little Too Late.

And the few things they did to try and make Zoe unique didn't have enough thought or effort put into them so they fall flat.

So she got bullied out of her old school because she "was playing a part; being someone else" and when she dropped the act, "everyone turned against me". Um, yeah, if I found out my friend who'd I'd been confiding in and getting close to lied and faked their entire personality, I wouldn't want to be friends with them anymore either. I wouldn't go so far as to put cockroaches in their locker, but I can't blame Zoe's old "friends" for feeling betrayed.

(And let's be real, the writers definitely HAD to throw in the cockroaches thing so that Zoe actually had examples of being bullied instead of "no one wanted to be my friend anymore when they found out I was a totally different person")

So she's an actress, putting on different faces to please the people around her (clearly, she learned nothing from her previous experience - it's reasonable that people didn't like you because you lied, not because they didn't like who you really are). But then they didn't even keep up this 'acting' thing for more than one episode! Sure, she dusts it off in "Gabriel Agreste", but that entire aspect of Zoe's story is resolved as soon as it's introduced.

(Though it's obvious that the show is trying to convince you that her old friends DID turn on her JUST because they didn't like her as a goody-two shoes nice girl).

With Zoe putting on a fake persona, there could've been intrigue over the episode, or even the season, about who Zoe really is. Like, Marinette is SURE that Zoe is faking being like Chloe, but she has no way to know Zoe wasn't faking being nice to her when they first met. Why not show Zoe putting on different personas than just those two? Have Zoe interact with Juleka by being low energy and stand offish, then swapping to bubbly and energetic around Rose? Have Marinette be confused and that becomes part of the Mystery of Zoe?

And how about that contrivance, that Marinette met "nice" Zoe first and so is SOOOOO sure that Zoe is faking being "mean"? And because Marinette declares this To Be So, all of her friends just believe Marinette and are willing to give Zoe a chance. (despite not believing her about other things *cough* Lila *cough*) This is the same Marinette who's canonically declared her hatred of "liars" more than once, so even if Zoe is only "pretending" to be mean, why is Marinette giving her a pass? I'll tell you why - because we need to seamlessly integrate Zoe into the main friend group so we can speed run her taking over as the new Bee.

Because that's the main problem with Zoe - it's so blatantly obvious what the writers are trying to get out of her, because they don't even TRY to hide it. Don't get it twisted, I don't care that Zoe is brought in to replace Queen Bee, Queen Bee deserves to be replaced. I care how Zoe was brought in to replace Queen Bee. If the writers had taken enough time and care to ACTUALLY seamlessly integrate Zoe, instead of just shoving her in and bending the characters to accept her with no hesitation, then Zoe wouldn't come across as such a Mary Sue-Diet Marinette.

We already have a pretty full cast, so making the audience care about this new character that they've neither hinted at or bothered to build up to in Season 4 was already a big ask. Which is why I believe they should've stretched her "faking personas" story over several episodes. Have Marinette be as wary about Zoe as she was about Lila, esPECIALLY since she's Chloe's sister, but then have Zoe showcase her range of "personas" to throw Marinette off guard. Then you can have scenes of Zoe's mask slipping, like small acts of heroism where there's no time to think: pulling someone out of the way of an akuma attack, distracting the akuma if it gets too close to Ladybug, or even using her acting to help someone else get out of trouble despite Zoe getting nothing in return. Show the REAL Zoe through her actions DESPITE her acting, which then justifies Marinette being sure that "mean" Zoe isn't the real one.

Could've even had Chloe notice that Zoe's slipping and isn't acting very much "like a Bourgeois" and make it CHLOE who digs up that Zoe was bullied, make it CHLOE who airs out all the things that were done to Zoe. I mean, who better to call out a poser than Chloe? Especially since the rest of the class would more willing to take Zoe's side against Chloe (you would think) and it helps with the whole Zoe vs Chloe thing with the Bee Miraculous. Plus, having that information aired out when Zoe didn't want it to could've been a better reason to be akumatized than just "I'm giving in to playing the role I was already determined to play and putting my shoes in this hole with my dreams." Just saying.

But they didn't want to take the time. They didn't want to try. They didn't want to put in the effort to make Zoe someone to care about, both for the audience and for the characters. They shoved her in, gave her an exposition dump, and suddenly everyone was totally cool with her, No Questions Asked. In fairness, they didn't actually have Zoe be mean to anyone other than Marinette One Time in front of Chloe, so there wasn't much to overcome. But that's just another check in the "wow, you didn't even try" column.

I can't bring myself to care about Zoe because she's just a tool. She was brought in with the sole purpose of putting in nearly no effort to become friends with everyone and use the Bee Miraculous, as quickly as possible and as effortlessly as possible. Never mind that such a thing probably should have been earned. I just hate it when I can see the writer's intentions so blatantly.

But I get it, the Bee Miraculous is seriously OP and it'd be a shame not to use it just because Queen Bee was benched. Too bad Ladybug can't just merge Miraculouses and use it hersel- ohhhh, wait.

[piratesandpizza]

Here’s my extremely belated secret Santa for @watsons-falling-snow !

A very romantic Adrienette kiss in front of the Arc de Triomphe (which I am quite proud of in this illustration lol)

I hope you enjoy this gift! and that your 2024 is amazing!

@mlsecretsanta

[hufflepotato-18]

i call this the “fast and furious” effect

[vizthedatum]

AI is not the problem—we are. It's the way people and corporations devalue the work of creators, the way data is gathered and trained for these generative models, and the way we look to these as solutions, forgetting that they are only tools (that could be used way more ethically).

“AI” is just math, tech, statistics, and computing power... Built upon the resources mined from this earth and the minds of our collective species.

I believe that our creative and natural resources can never be replaced in their worth.

I am a data scientist by trade, and I build statistical models and AI. I use (non-generative) AI to help me understand how the world works and make decisions (including creation).

I sometimes program my art (or I use tools like Inkscape or similar software to help me! I use a computer!!) - because I can write instructions like a recipe where I gather all my ingredients with consent and fair use. (I can program, is what I’m saying. I can curate what data I'm using.)

I can have a tool to carry out my will to create. I am not arguing about that.

So, I understand the disability argument being made with generative AI.

Accessibility is important, and I agree with making information and creation more accessible.

I love that people can learn about stuff. My suggestion is for groups in power to make better search engines, tutorials, and tools for things with AI. Help people with formats! Help people automate!! I don't think there's anything wrong with that.

—

But here is the nuance to this:

—

Generative AI is an unethical tool if it feeds off work you can't see or validate.

This a corrupted data issue. This is a corrupted “people in power” issue.

There is a lack of transparency in these tools' decision-making. AI's intelligence is only as good as the intelligence of the data and algorithms.

It is a regurgitation of information that is feeding on itself.

So I don’t really care about people who use tools to create things - but I do care when those outputs are not crediting other artists’ works. I do care when systemic discrimination shows up in AI results because our data is inherently flawed (because our data is from our people).

We must be more critical and nuanced about how we talk about this.

The technology (statistical, machine, and deep learning models) has been around for a long time. Papers on deep learning have existed for decades. As a researcher, I find the math fascinating.

The leaps have been tremendous, but again… it’s not the nature of the tools themselves.

It’s us.

How are we using these tools, what information is being given to these tools, and what are the lasting impacts on our society?

[wooawi]

Take Two Chapter 14

preview under the cut

Plagg had woken up, the noise too much for him to ignore like usual, and Tikki cursed the fragility of humans. Surely they didn’t have to scream every time.

“Acting like what, Plagg?” She said and lowered her paw. Alya’s body released and the girl dropped to the floor, dry heaving and still sobbing.

Plagg yawned and floated closer to her, his drooping eyes taking in the scene around him. “Like this, sugarcube. If I had known you would get this petty, I would’ve stopped you when Nooroo approached me all those millenia ago.”

She wasn’t surprised that the other kwami had been talking about her behind her back. They couldn’t rationalise why she did the things she did. The burden of keeping the universe balanced was something they could never understand, especially when you have to carry your own weight and the weight of your partner because he would rather spend his lifetime eating cheese and being lazy and–

“Your thinking is so loud,” Plagg complained and rolled over in the air. His ears perked up when Alya hiccuped through her tears, and he groaned and made his way to her. One touch was all it took, and the effects of Tikki’s rage was erased completely. The girl sat up straight, shocked and confused, and tensed herself up for more pain. “All you have to do is fix your mistakes and we can all be done with this and go back.”

Sometimes, she truly hated Plagg. She alone had been carrying the reputation of the kwami on her back for all this time with nothing but contempt and disgust from the others, and now that the cat kwami was finally awake and coherent enough to look around, he decided to fight against her?

“Uh oh, I know that look,” he said and floated back to his Chosen. “You’re doing the whole ‘Tikki vs The World’ thing again, aren’t you.”

“Are you capable of taking anything seriously, you stupid excuse for a god?”

Plagg just grinned.

[rosekasa]

kofi comm for @blackholesoffandom !!! as per usual katie always just pays for me to be happy shdfjkdsfds <3 <3 <3 thank you so much as usual! i love u dearly!!

kofi sketch comms open!

[katydoodles]

Planning Session

Chloe and Zoe trying to figure out how to take the credit card again.

[etanow]

It’s SPOOKY MONTH ya dead girl 🧟

[songship]

I get mad when I see anyone talking about ML/AI on the internet ever because it makes me so incandescently mad that it's being used wrong because there are genuinely really good applications!! And all people want to do is to fucking replace people's jobs but doing it worse or shove it into random shit!!! And it also makes me mad that it maligns all the existing cool and good projects that get painted with the same tainted brush so anything ML ever gets slapped with the "AI bad" label automatically out of reflex because all the bullshit!!! Aughhhhh

[jcmarchi]

Peter Ellman, President and CEO of Certis Oncology Solutions – Interview Series

New Post has been published on https://thedigitalinsider.com/peter-ellman-president-and-ceo-of-certis-oncology-solutions-interview-series/

Peter Ellman, President and CEO of Certis Oncology Solutions – Interview Series

Certis Oncology Solutions, led by Peter Ellman, President and CEO, is a life science technology company dedicated to realizing the promise of precision oncology. The company’s product is Oncology Intelligence® — highly predictive therapeutic response data derived from advanced cancer models. Certis partners with physician-scientists and industry researchers to expand access to precision oncology and address the critical translation gap between preclinical studies and clinical trials.

Can you describe the broader problem in oncology research that the CertisOI Assistant is addressing?

The failure rate of oncology investigational drug candidates is high. It was recently reported that in 2023, 90% of oncology programs ultimately failed. That figure is a remarkable improvement over the historical trend, which hovered around 96% until 2022. Considering the cost of developing drugs, a 90% failure rate is not sustainable. Imagine how patients would benefit if the success rate were even 50%.

CertisOI Assistant immediately addresses two really important issues that contribute to this failure rate:

Improved preclinical model selection: Many compounds show promising results in preclinical studies but fail to demonstrate a sufficient therapeutic effect in humans.Most members of the scientific community point to preclinical models as part of the problem. Choosing preclinical models with the correct gene expression signature (and using orthotopic engraftments for pivotal studies) can improve “translation” into the clinic.

Earlier, better biomarker identification: Relying on biomarkers that do not accurately predict therapeutic response can result in failed clinical trials. CertisOI Assistant is integrated with CertisAI, our patent-pending predictive AI/ML platform, enabling the identification of predictive biomarkers early in the drug development process.

How does the CertisOI Assistant use AI to improve access to oncology data, and what sets it apart from other AI tools in the field?

The CertisOI Assistant provides advanced data analysis and predictive modeling capabilities through an easy-to-use, natural language interface. It stands out in several ways:

Comprehensive Dataset Integration: The assistant integrates a wide range of oncology data, including patient information, tumor characteristics, genetic profiles, and drug response predictions. This holistic approach allows for a more comprehensive analysis than tools focusing on isolated data types.

AI-Based Predictions: The assistant employs AI algorithms to predict drug response and resistance, offering insights into which treatments will likely be effective for specific cancer models. This predictive capability is crucial for personalized medicine and sets it apart from tools that rely solely on historical data.

User-Friendly Interface: By providing an intuitive interface for querying and analyzing complex datasets, the assistant makes it easier for researchers to access and interpret oncology data without requiring advanced technical skills.

Focus on Pre-Clinical Models: The assistant specializes in pre-clinical oncology research, particularly PDX and cell line models, offering unique insights into early-stage drug development and tumor biology.

Interactive Visualizations: The assistant supports interactive visualizations, such as pharmacology and tumor growth studies, enabling researchers to explore data more engaging and informatively.

How does the tool transform complex data into actionable insights, especially for researchers working on drug sensitivity or genomic data?

CertisOI Assistant leverages a structured workflow to transform raw data into meaningful insights. It involves querying a comprehensive oncology dataset, analyzing the data, and presenting the results in a clear and interpretable format. Here’s how it works:

Data Querying: CertisOI Assistant can access a relational database containing detailed information about oncology models, including patient data, tumor characteristics, genomic data, and drug response predictions. It uses SQL-like queries to extract relevant data based on the researcher’s specific needs.

Data Analysis: Once the data is retrieved, CertisOI Assistant can perform various analyses, such as identifying common mutations, correlating gene expression with drug sensitivity, or evaluating pharmacology study results. It can also rank and filter data to highlight the most significant findings.

Visualization: The assistant can present data in tabular formats, generate interactive charts for pharmacology and tumor growth studies, and display histology images. This visualization helps researchers quickly grasp complex data patterns and relationships.

Interpretation and Insights: By providing a clear interpretation of the data, including predictions for drug sensitivity or resistance, CertisOI Assistant helps researchers make informed decisions about potential therapeutic strategies or further experimental directions.

Customization and Flexibility: Researchers can tailor their queries to focus on specific cancer types, genetic markers, or treatment responses, allowing for a highly customized analysis that aligns with their research objectives.

How does the CertisOI Assistant enhance researchers’ ability to select cancer models, design biomarker strategies, or perform in silico validations?

I covered the first two areas – the cancer model section and biomarker strategy design – at the outset of this interview, so I’ll focus on performing in silico validations. CertisOI Assistant provides a virtual environment to test and validate hypotheses related to drug efficacy, target engagement, and biomarker discovery without the need for immediate laboratory experiments. This allows them to rapidly refine their hypotheses and focus experimental efforts on the most promising avenues.

Here are a few examples:

Drug Response Predictions: Use AI-based predictions for drug response and resistance to assess how different models are likely to respond to specific drugs. This can help validate the potential efficacy of a drug in silico before moving to in vitro or in vivo studies.

Genomic and Molecular Profiling: Analyze the genomic data, including mutations, gene expression, and copy number variations, to identify potential targets and validate their relevance to the drug’s mechanism of action. This can help in understanding the molecular basis of drug sensitivity or resistance.

Biomarker Discovery: Correlate molecular characteristics with drug response predictions to identify potential predictive biomarkers. This can guide the selection of patient populations more likely to benefit from a particular therapy.

Combination Therapy Exploration: Explore drug synergy predictions to identify promising drug combinations that may enhance therapeutic outcomes. This can provide insights into potential combination strategies that can be further validated experimentally.

Histological Analysis: Use histology images to validate the morphological effects of drugs on tumor tissues, providing additional evidence for the drug’s mechanism of action and potential efficacy.

Cross-Model Comparisons: Compare different models to understand how various genetic backgrounds influence drug response, helping to validate hypotheses about the role of specific genes or pathways in silico.

Virtual Screening: Perform virtual screening of drugs against a wide range of models to prioritize candidates for further experimental validation.

Can you share examples of how researchers are anticipated to use this tool to improve their workflows or achieve breakthroughs?

The simplest example is preclinical model selection. Every preclinical study begins with the selection of tumor models. CertisOI Assistant takes the manual effort out of this process and brings great precision to selecting the optimal models for any given study.

Another is developing a biomarker strategy. The traditional approach is to hypothesize what biomarker or biomarkers might be linked to the drug’s mechanism of action and then test those hypotheses in preclinical studies, which is usually an iterative process. If preclinical data is promising, researchers must validate predictive biomarkers in human clinical trials—and as discussed, the failure rate is high.

The CertisOI Assistant helps researchers identify and validate more precise, predictive gene expression biomarkers earlier in the development process and with less iteration than the traditional workflow—saving time, and money, and improving chances for commercial success.

What kinds of cancer models or datasets does the tool support, and how does this breadth benefit the research community?

The current version of CertisOI gives researchers access to Certis’ rapidly expanding library of PDX and PDX-derived tumor models and the entire Cancer Cell Line Encyclopedia (CCLE) of models. The platform’s algorithms also draw on data from Genomics of Drug Sensitivity in Cancer (GDSC), International Cancer Genome Consortium (ICGC), CI ALMANAC, O’Neil, and other datasets. This holistic approach to data integration allows for a more comprehensive analysis than tools that focus on isolated data types.

The CertisOI Assistant is designed to be user-friendly. How do you ensure that it is accessible to researchers who may not have extensive technical expertise?

Several features make CertisOI Assistant accessible to researchers at all levels:

Intuitive Interface: The interface is designed to be intuitive and easy to navigate, allowing users to perform complex queries and analyses without needing to understand the underlying technical details.

Guided Workflows: The assistant provides guided workflows for common research tasks, such as querying drug response predictions, analyzing genomic data, and exploring pharmacology studies. This helps users focus on their research questions without getting bogged down in technical complexities.

Natural Language Processing: Users can interact with the assistant using natural language queries, making accessing the information they need easier for those without technical expertise. The assistant interprets the queries and translates them into the appropriate database queries.

Comprehensive Documentation: Detailed documentation and tutorials help users understand how to use the assistant effectively. This includes step-by-step guides, examples, and explanations of key concepts.

Interactive Visualizations: The assistant provides interactive visualizations for data analysis, such as charts and histology images, allowing users to explore and interpret data visually without needing to write code.

Responsive Support: Users can access responsive support to assist with any questions or issues. This ensures they can get help quickly and continue their research without unnecessary delays.

Customizable Queries: While the assistant provides default workflows, it also allows for customization, enabling users to tailor queries to their specific research needs without requiring deep technical knowledge.

Collaboration is a key aspect of research. How does the CertisOI Assistant facilitate teamwork among researchers or institutions?

With CertisOI Assistant, researchers from different teams or institutions can access the same dataset and tools, allowing them to work collaboratively on shared projects or research questions. The platform also makes it easy to download and share data queries, results, and insights among team members so everyone involved in a project can contribute effectively.

What are the biggest challenges in scaling AI adoption in cancer research, and how can they be addressed?

Significant challenges include data security, data integration, and trust in AI‐based outcome predictions. I am not an expert on data security or data integration, but great minds are working to solve those challenges. With respect to trusting AI-generated predictions, we need efficient and credible ways to validate those predictions.

Certis has taken a two-pronged approach to this: in silico validation via internal, cross-validation studies, and in vivo validation—performing studies in clinically relevant mouse models to evaluate the accuracy of our platform’s predictions. Over time, these tools will also be validated clinically in human patients—but of course, that will take a great deal of time and money, as well as the willingness to change the current cancer treatment paradigm. The medical and regulatory community will have to stop relying on how things have always been done and embrace the power of computational analyses to inform decisions.

How do you envision tools like the CertisOI Assistant shaping the future of cancer treatment and precision medicine?

Modern medicine doesn’t yet have a great way to match patients to the ideal treatments. Overall, only 10% of cancer patients experience a clinical benefit from treatments matched to tumor DNA mutations. That not only hurts patients’ health, but it also harms them financially. An estimated $2.5 billion —with a B—is wasted on ineffective therapies. It is a very sad fact that 42% of cancer patients fully deplete their assets by the second year of their diagnosis.

Tools like CertisOI Assistant and CertisAI will help us realize the promise of precision medicine—getting people the optimal treatment for their unique form of cancer the first time, every time…. And to democratize access to more effective, personalized care.

Thank you for the great interview, readers who wish to learn more should visit Certis Oncology Solutions.