France has announced a new national AI strategy, including government funding worth nearly €1.5 billion ($1.85 billion). But the country's president, Emmanuel Macron, is worried about the damage this technology could do if not properly guided. In an interview with Wired, he said there was even a risk AI could "jeopardize democracy." Macron is worried about unaccountable "black box" algorithms being introduced into society and making decisions formerly entrusted to humans. He gave the example of an algorithm used to sort students into universities and said that if its workings were not easy to understand, it could destroy trust and encourage people to "reject" innovation.

Artificial intelligence has increasingly been integrated into the weapons systems of the world's leading militaries, and at least one expert has said the futuristic technology may soon be the subject of a new Cold War. In a piece published Tuesday by The Conversation, North Dakota State University assistant professor Jeremy Straub argued that unlike the nuclear weapons that dominated much of the 21st century arms race between the U.S. and the Soviet Union, the use of cyberweapons and artificial intelligence largely remained "fair game," even as tensions again flared between the rivals. Both countries have invested heavily in developing new tools to wage war on this new front, but Russia particularly has sought to use it as an opportunity to upstage the more conventionally powerful U.S. Related: U.S. is losing to Russia and China in war for artificial intelligence, report says "Now, more than 30 years after the end of the Cold War, the U.S. and Russia have decommissioned tens of thousands of nuclear weapons. Any modern-day cold war would include cyberattacks and nuclear powers' involvement in allies' conflicts," wrote Straub, who was also associate director of the university's Institute for Cyber Security Education and Research, in his article. "It's already happening," he added.

Munich: Fujitsu has embarked on a long-term research and co-creation program with the French National Institute for Research in Computer Science and Automation (Inria). Just one year after the start of their partnership, the two organizations have formally committed to even closer collaboration, reflecting Fujitsu's commitment to driving digital innovation in France. This new program combines Inria's expertise in AI-focused research and development with Fujitsu's technology. A joint team comprising engineers from Fujitsu in Japan and Inria will work closely together, focused on developing new Artificial Intelligence and machine learning techniques by leveraging advanced mathematics and computing. Artificial intelligence will be deployed to interpret IoT data, to generate insights for customers.

I recently had the pleasure of being invited to speak at The Item Conference http://www.item.nu/cgi-oic/pagedb.exe/show?no 1 in London for educators, policy-makers and head-teachers visiting our amazing City for inspiration and knowledge about how to foster creativity in children with I. T. They were particularly interested in the possibilities of Artificial Intelligence and Machine Learning. So, with thinking cap on, and just a few short hours to prepare, I was thrilled to find out that A.I in education is not the work of science fiction, but is with us right now -- in action, and starting to build impact. For hundreds of years, humans have pondered the idea of building intelligent machines. Over this time, artificial intelligence has had highs and lows, demonstrated successes and unfulfilled potential. Today, the news is filled with the application of AI and machine learning to new problems.

French President Emmanuel Macron announced his intention to make France an AI leader and avoid "dystopia", supported by €1.5bn in investment. France is not the first country to surge ambitiously towards establishing itself as a "leader" in AI. Putin famously stated that "whoever becomes the leader in this sphere will become the ruler of the world". China has a three-year action plan to establish itself at the top. Canada's Trudeau discussed on multiple occasions the consequences of automation, and the opportunities of artificial intelligence.

We prove a \emph{query complexity} lower bound for approximating the top $r$ dimensional eigenspace of a matrix. We consider an oracle model where, given a symmetric matrix $\mathbf{M} \in \mathbb{R}^{d \times d}$, an algorithm $\mathsf{Alg}$ is allowed to make $\mathsf{T}$ exact queries of the form $\mathsf{w}^{(i)} = \mathbf{M} \mathsf{v}^{(i)}$ for $i$ in $\{1,...,\mathsf{T}\}$, where $\mathsf{v}^{(i)}$ is drawn from a distribution which depends arbitrarily on the past queries and measurements $\{\mathsf{v}^{(j)},\mathsf{w}^{(i)}\}_{1 \le j \le i-1}$. We show that for every $\mathtt{gap} \in (0,1/2]$, there exists a distribution over matrices $\mathbf{M}$ for which 1) $\mathrm{gap}_r(\mathbf{M}) = \Omega(\mathtt{gap})$ (where $\mathrm{gap}_r(\mathbf{M})$ is the normalized gap between the $r$ and $r+1$-st largest-magnitude eigenvector of $\mathbf{M}$), and 2) any algorithm $\mathsf{Alg}$ which takes fewer than $\mathrm{const} \times \frac{r \log d}{\sqrt{\mathtt{gap}}}$ queries fails (with overwhelming probability) to identity a matrix $\widehat{\mathsf{V}} \in \mathbb{R}^{d \times r}$ with orthonormal columns for which $\langle \widehat{\mathsf{V}}, \mathbf{M} \widehat{\mathsf{V}}\rangle \ge (1 - \mathrm{const} \times \mathtt{gap})\sum_{i=1}^r \lambda_i(\mathbf{M})$. Our bound requires only that $d$ is a small polynomial in $1/\mathtt{gap}$ and $r$, and matches the upper bounds of Musco and Musco '15. Moreover, it establishes a strict separation between convex optimization and \emph{randomized}, "strict-saddle" non-convex optimization of which PCA is a canonical example: in the former, first-order methods can have dimension-free iteration complexity, whereas in PCA, the iteration complexity of gradient-based methods must necessarily grow with the dimension.

A problem of paramount importance in both pure (Restricted Invertibility problem) and applied mathematics (Feature extraction) is the one of selecting a submatrix of a given matrix, such that this subma-trix has its smallest singular value above a specified level. Such problems can be addressed using perturbation analysis. In this paper, we propose a perturbation bound for the smallest singular value of a given matrix after appending a column, under the assumption that its initial coherence is not large, and we use this bound to derive a fast algorithm for feature extraction.

Our usage of language is not solely reliant on cognition but is arguably determined by myriad external factors leading to a global variability of linguistic patterns. This issue, which lies at the core of sociolinguistics and is backed by many small-scale studies on face-to-face communication, is addressed here by constructing a dataset combining the largest French Twitter corpus to date with detailed socioeconomic maps obtained from national census in France. We show how key linguistic variables measured in individual Twitter streams depend on factors like socioeconomic status, location, time, and the social network of individuals. We found that (i) people of higher socioeconomic status, active to a greater degree during the daytime, use a more standard language; (ii) the southern part of the country is more prone to use more standard language than the northern one, while locally the used variety or dialect is determined by the spatial distribution of socioeconomic status; and (iii) individuals connected in the social network are closer linguistically than disconnected ones, even after the effects of status homophily have been removed. Our results inform sociolinguistic theory and may inspire novel learning methods for the inference of socioeconomic status of people from the way they tweet.

This paper proposes a method for generating speech from filterbank mel frequency cepstral coefficients (MFCC), which are widely used in speech applications, such as ASR, but are generally considered unusable for speech synthesis. First, we predict fundamental frequency and voicing information from MFCCs with an autoregressive recurrent neural net. Second, the spectral envelope information contained in MFCCs is converted to all-pole filters, and a pitch-synchronous excitation model matched to these filters is trained. Finally, we introduce a generative adversarial network -based noise model to add a realistic high-frequency stochastic component to the modeled excitation signal. The results show that high quality speech reconstruction can be obtained, given only MFCC information at test time.

This paper studies the complexity of the stochastic gradient algorithm for PCA when the data are observed in a streaming setting. We also propose an online approach for selecting the learning rate. Simulation experiments confirm the practical relevance of the plain stochastic gradient approach and that drastic improvements can be achieved by learning the learning rate.