It has become increasingly common to pre-train models to develop general-purpose abilities and knowledge that can then be “transferred” to downstream tasks.
In applications of transfer learning to computer vision, pre-training is typically done via supervised learning on a large labelled dataset like ImageNet. In contrast, modern techniques for transfer learning in NLP often pre-train using unsupervised learning on unlabeled data.
In spite of being widely popular there are still few pressing questions bothering transfer learning in ML. Read More
Monthly Archives: July 2020
Kai-Fu Lee Gives AI a B-Minus Grade in the Covid-19 Fight
Robots and computer programs can help with social distancing and food delivery, but have been less helpful in developing a vaccine.
This past week, as part of the Aspen Ideas Festival, I spoke with Kai-Fu Lee, the president and chair of Sinovation Ventures and a pioneer in artificial intelligence. We discussed his recent argument that AI has been of limited use in the response to the coronavirus crisis. And then we talked about the future of work and why he thinks that Covid-19 is going to accelerate trends toward automation. Because of the virus, and because of the way we all work now, we’re going to have many more robots and other machines in our factories, restaurants, and kitchens. A lightly edited transcript is below. You can watch the original video here. Read More
5G will change the world. China wants to lead the way
China isn’t the only country jockeying for control. The US dominated 4G’s expansion and expects to do the same now.
China is moving full steam ahead on 5G, barely slowed down by a pandemic that has ravaged the world. This is setting up a race between the nation and the US, which led the way with 4G cellular technology and wants to keep its pole position in this next generation. Read More
Augmented Intelligence is the New Intelligence
What does it mean when shifting from AI to augmented intelligence?
The future of decision-making includes an inventive blend of information, analytics, and artificial intelligence (AI), with the perfect scramble of human judgment. The outcome is augmented intelligence, where the analytical force and speed of AI assumes control over most of data processing, controlling human workers to make progressively agile, more intelligent choices and find new discoveries. Read More
The path to real-world artificial intelligence
Experts from MIT and IBM held a webinar this week to discuss where AI technologies are today and advances that will help make their usage more practical and widespread. Read More
Improving Robustness of Deep-Learning-Based Image Reconstruction
Deep-learning-based methods for different applications have been shown vulnerable to adversarial examples. These examples make deployment of such models in safety-critical tasks questionable.Use of deep neural networks as inverse problem solvers has generated much excitement for medical imaging including CT and MRI, but recently a similar vulnerability has also been demonstrated for these tasks. We show that for such inverse problem solvers, one should analyze and study the effect of adversaries in the measurement-space,instead of the signal-space as in previous work. Read More
Does Adversarial Transferability Indicate Knowledge Transferability?
Despite the immense success that deep neural networks (DNNs) have achieved,adversarial examples, which are perturbed inputs that aim to mislead DNNs to make mistakes, have recently led to great concern. On the other hand, adversarial examples exhibit interesting phenomena, such as adversarial transferability. DNNs also exhibit knowledge transfer, which is critical to improving learning efficiency and learning in domains that lack high-quality training data. In this paper, we aim to turn the existence and pervasiveness of adversarial examples into an advantage.Given that adversarial transferability is easy to measure while it can be challenging to estimate the effectiveness of knowledge transfer,does adversarial transferability indicate knowledge transferability? We first theoretically analyze the relationship between adversarial transferability and knowledge transferability and outline easily checkable sufficient conditions that identify when adversarial transferability indicates knowledge transferability. In particular, we show that composition with an affine function is sufficient to reduce the difference between two models when adversarial transferability between them is high. We provide empirical evaluation for different transfer learning scenarios on diverse datasets, including CIFAR-10,STL-10, CelebA, and Taskonomy-data – showing a strong positive correlation be-tween the adversarial transferability and knowledge transferability, thus illustrating that our theoretical insights are predictive of practice. Read More
#adversarialSystematic Literature Review to Investigate the Application of Open Source Intelligence (OSINT) with Artificial Intelligence
Open Source Intelligence (OSINT) is a concept to describe the search, collection, analysis, and use of information from open sources, as well as the techniques and tools used. OSINT emerges out of a military need to collect relevant and publicly available information. Through the use of OSINT, it is possible to find specific information that has some knowledge or provides an advantage. Since its emergence, some studies have been done proposing and developing new ways of using OSINT in different areas. In addition to OSINT, another field of study that has also been a worldwide trend and is being used together with other areas is Artificial Intelligence (AI). AI is the area of computer science responsible for the development of intelligent systems. However, a systematic literature review that investigates the use of OSINT over the years and your application with AI was not found. So, this work has an objective to develop a systematic literature review on OSINT to investigate the application of OSINT with AI. Read More
#icHow AI can empower communities and strengthen democracy
Each Fourth of July for the past five years I’ve written about AI with the potential to positively impact democratic societies. I return to this question in hopes of shining a light on technology that can strengthen communities, protect privacy and freedoms, and otherwise support the public good.
This series is grounded in the principle that artificial intelligence is capable of not just value extraction, but individual and societal empowerment. While AI solutions often propagate bias, they can also be used to detect that bias. As Dr. Safiya Noble has pointed out, artificial intelligence is one of the critical human rights issues of our lifetimes. AI literacy is also, as Microsoft CTO Kevin Scott asserted, a critical part of being an informed citizen in the 21st century. Read More
The US, China and the AI arms race: Cutting through the hype
The reality is that US and China efforts to develop AI are entwined, even if the tensions of coronavirus and trade disagreements may spur a separation.
… the narrative of an AI “arms race” between the US and China has been brewing for years. Dramatic headlines suggest that China is poised to take the lead in AI research and use, due to its national plan for AI domination and the billions of dollars the government has invested in the field, compared with the US’ focus on private-sector development.
But the reality is that at least until the past year or so, the two nations have been largely interdependent when it comes to this technology. It’s an area that has drawn attention and investment from major tech heavy hitters on both sides of the Pacific, including Apple, Google and Facebook in the US and SenseTime, Megvii and YITU Technology in China. Read More
Rick's Cafe AI 