Berlin's annual IFA electronics show isn't typically a hotspot of PC news, since it follows hot on the heels of the computer-centric Computex in June. Intel used IFA to launch its hotly anticipated "Lunar Lake" Core Series Ultra 2 processors, packing a radical new architecture and seriously improved graphics chops. Not to be outgunned, Qualcomm revealed new 8-core variants of its Snapdragon X PC chips, aiming to bring multi-day battery life to sub- 900 laptops. With so many notebook reveals crammed into such a short time span, however, the deluge felt a bit same-y at times – but not with the delightfully unorthodox systems below, which lean on surprisingly nifty gimmicks to stand out from the crowd. Without further ado, let's get weird.

Alignment training is crucial for enabling large language models (LLMs) to cater to human intentions and preferences. It is typically performed based on two stages with different objectives: instruction-following alignment and human-preference alignment. However, aligning LLMs with these objectives in sequence suffers from an inherent problem: the objectives may conflict, and the LLMs cannot guarantee to simultaneously align with the instructions and human preferences well. To response to these, in this work, we propose a Hybrid Alignment Training (Hbat) approach, based on alternating alignment and modified elastic weight consolidation methods. The basic idea is to alternate between different objectives during alignment training, so that better collaboration can be achieved between the two alignment tasks.We experiment with Hbat on summarization and dialogue tasks. Experimental results show that the proposed \textsc{Hbat} can significantly outperform all baselines. Notably, Hbat yields consistent performance gains over the traditional two-stage alignment training when using both proximal policy optimization and direct preference optimization.

We present a new technique for time series analysis based on dy(cid:173) namic probabilistic networks. In this approach, the observed data are modeled in terms of unobserved, mutually independent factors, as in the recently introduced technique of Independent Factor Anal(cid:173) ysis (IFA). However, unlike in IFA, the factors are not Li.d.; each factor has its own temporal statistical characteristics. We derive a family of EM algorithms that learn the structure of the underlying factors and their relation to the data. These algorithms perform source separation and noise reduction in an integrated manner, and demonstrate superior performance compared to IFA.

Warren Vickers is a man who has always marched to the beat of his own drum – quite literally.

Bayesian nonparametrics based on completely random measures (CRMs) offers a flexible modeling approach when the number of clusters or latent components in a dataset is unknown. However, managing the infinite dimensionality of CRMs often leads to slow computation. Practical inference typically relies on either integrating out the infinite-dimensional parameter or using a finite approximation: a truncated finite approximation (TFA) or an independent finite approximation (IFA). The atom weights of TFAs are constructed sequentially, while the atoms of IFAs are independent, which (1) make them well-suited for parallel and distributed computation and (2) facilitates more convenient inference schemes. While IFAs have been developed in certain special cases in the past, there has not yet been a general template for construction or a systematic comparison to TFAs. We show how to construct IFAs for approximating distributions in a large family of CRMs, encompassing all those typically used in practice. We quantify the approximation error between IFAs and the target nonparametric prior, and prove that, in the worst-case, TFAs provide more component-efficient approximations than IFAs. However, in experiments on image denoising and topic modeling tasks with real data, we find that the error of Bayesian approximation methods overwhelms any finite approximation error, and IFAs perform very similarly to TFAs.

Of all the annual meetings of tech-obsessed minds, IFA is perhaps the most unpredictable. You know everyone and their mother is going to CES to set the tone for the year, and MWC is all things mobile. At IFA, though, you never quite know what surprises lurk behind the floor-to-ceiling displays of kitchen appliances that dominate the halls of the sprawling Messe Berlin conference center. History and rumors give us some idea, however. Naturally, we'll be on the ground to bring you all the important news, but these are some of things we expect to see announced at this year's show.

Ahead of IFA world's leading trade show for consumer electronics and home appliance slated to hold this August in Berlin, Germany, LG Electronics Chief Executive Officer, Jo Seong-Jin and Chief Technology Officer, Dr. I.P. Park will deliver a joint opening keynote at the trade exhibition to outline their vision for LG's ThinQ strategy for Artificial Intelligence. The first IFA keynote for both executives, titled Think Wise, Be Free: Living Freer with AI will be delivered at the trade expo. With interest in all things AI at a peak, CEO Jo, according to LG Electronics would provide his insight into how LG's AI strategy will change customers' lives based on its unique philosophy of an "open platform, open partnership and open connectivity". Similarly, Dr. Park, according to the technology company is expected to show how this technology comes to life in everyday products like refrigerators, TVs and washing machines, and how LG's open AI strategy benefits consumers, thanks to its focus of giving them both control and convenience. "The IFA keynotes provide CEOs and top executives with a global media platform for forward thinking ideas and strategies," Jens Heithecker, Messe Berlin Group executive vice president and IFA executive director, said, adding that, "LG has been and continues to be a key global player in the field of artificial intelligence, which is why we are extremely pleased that Mr. Jo and Dr. Park will deliver the opening keynote at IFA this August."

Artificial intelligence is more than just a buzzword; it's an inevitable reality. While robots might not walk, talk and live among us, through automation technology they are still making their way into our lives, streamlining everything from our day-to-day routines to our business activities. Robo advisors are the new craze in the world of personal finance and investment… Okay, the term might be misleading as they really have nothing to do with robots – at least not in the traditional sense. However, they certainly emphasise a direction of travel, one which demonstrates just how powerful automation technology is becoming. Robo advisors are online investment platforms designed to take clients through the investment process automatically, so that independent financial advisors (IFA) can be bypassed.

As we wind down after trawling IFA's labyrinthine halls, covering everything from phones to washing machines, wearables to haunted pianos, we wanted to point out the most notable things to come out of Europe's biggest tech show -- and it was quite the show this year. If you don't read anything else, read this. And if you do want even more, you can find everything else right here. When Microsoft began talking about its Mixed Reality program earlier this year, the whole thing left me pretty cold. The company claimed that it could slash the cost of a VR headset almost in half, and that the kit would even work with PCs that used only integrated graphics.

Mark Zuckerberg made it his new year's goal to create Jarvis-like artificial intelligence for his home -- at this point it lets him manage his house and toasts his bread. But a German startup called BuddyGuard may have beaten him to the punch. The company recently launched an artificially intelligent security system called Flare, and brought it to IFA to show it off. Flare is part camera, part AI, part IoT accessory. BuddyGuard bills its product as "the first home security system powered by true artificial intelligence."